Describe the bug
Cube fails to generate correct SQL when a join's ON condition relies on columns from other cubes (let's call them "intermediate cubes"), which are themselves joined to the primary query cube. If no fields are explicitly selected from these intermediate cubes, they are not included in the FROM clause, even though they are necessary for a subsequent join's ON condition. This results in an SQL error because the aliases/tables referenced in the final join's ON clause are not available.

In the example below, test_facts joins to merchant_and_product_dims. This join's ON condition ({merchant_dims.merchant_id} = {merchant_and_product_dims.merchant_id} AND {product_dims.product_id} = {merchant_and_product_dims.product_id}) requires merchant_dims and product_dims to be available. However, if no dimensions/measures are selected directly from merchant_dims or product_dims, Cube omits them from the FROM clause, breaking the join to merchant_and_product_dims.

To Reproduce

1. Define the following Cube schema:

   cubes:
     - name: merchant_dims
       sql: |
         (
           SELECT 101 AS merchant_sk, 'M1' AS merchant_id
           UNION ALL
           SELECT 102 AS merchant_sk, 'M2' AS merchant_id
         )
       public: false
       dimensions:
         - name: merchant_sk
           sql: merchant_sk
           type: number
           primary_key: true
         - name: merchant_id
           sql: merchant_id
           type: string
   
     - name: product_dims
       sql: |
         (
           SELECT 201 AS product_sk, 'P1' AS product_id
           UNION ALL
           SELECT 202 AS product_sk, 'P2' AS product_id
         )
       public: false
       dimensions:
         - name: product_sk
           sql: product_sk
           type: number
           primary_key: true
         - name: product_id
           sql: product_id
           type: string
   
     - name: merchant_and_product_dims
       sql: |
         (
           SELECT 'M1' AS merchant_id, 'P1' AS product_id, 'Organic' AS acquisition_channel
           UNION ALL
           SELECT 'M1' AS merchant_id, 'P2' AS product_id, 'Paid' AS acquisition_channel
           UNION ALL
           SELECT 'M2' AS merchant_id, 'P1' AS product_id, 'Referral' AS acquisition_channel
         )
       public: false
       dimensions:
         - name: product_id
           sql: product_id
           type: string
           primary_key: true
         - name: merchant_id
           sql: merchant_id
           type: string
           primary_key: true
         - name: acquisition_channel
           sql: acquisition_channel
           type: string
   
     - name: test_facts
       sql: |
         (
           SELECT DATE '2023-01-01' AS reporting_date, 101 AS merchant_sk, 201 AS product_sk, 100 AS amount
           UNION ALL
           SELECT DATE '2023-01-01' AS reporting_date, 101 AS merchant_sk, 202 AS product_sk, 150 AS amount
           UNION ALL
           SELECT DATE '2023-01-02' AS reporting_date, 102 AS merchant_sk, 201 AS product_sk, 200 AS amount
         )
       public: false
       joins:
         - name: merchant_dims
           relationship: many_to_one
           sql: "{CUBE}.merchant_sk = {merchant_dims.merchant_sk}"
         - name: product_dims
           relationship: many_to_one
           sql: "{CUBE}.product_sk = {product_dims.product_sk}"
         - name: merchant_and_product_dims # This join depends on merchant_dims and product_dims
           relationship: many_to_one
           sql: "{merchant_dims.merchant_id} = {merchant_and_product_dims.merchant_id} AND {product_dims.product_id} = {merchant_and_product_dims.product_id}"
       dimensions:
         - name: reporting_date
           sql: reporting_date
           type: time
           primary_key: true
         - name: merchant_sk
           sql: merchant_sk
           type: number
           primary_key: true
         - name: product_sk
           sql: product_sk
           type: number
           primary_key: true
         - name: acquisition_channel # This dimension triggers the join to merchant_and_product_dims
           sql: "{merchant_and_product_dims.acquisition_channel}"
           type: string
       measures:
         - name: amount_sum
           sql: amount
           type: sum

2. Execute a query that selects the acquisition_channel from test_facts (which implicitly requires joining merchant_and_product_dims) and any measure. For example:

   {
     "measures": ["test_facts.amount_sum"],
     "dimensions": [
       "test_facts.reporting_date",
       "test_facts.merchant_sk",
       "test_facts.product_sk",
       "test_facts.acquisition_channel"
     ]
   }

Actual Results:
The generated SQL is incorrect and causes an error as the "merchant_dims".merchant_id and "product_dims".product_id columns can not be resolved.

This is because merchant_dims and product_dims tables are referenced in the ON clause for merchant_and_product_dims but are not included in the FROM clause.

-- Incorrect SQL Generated
SELECT
  "test_facts".reporting_date,
  "test_facts".merchant_sk AS merchant_sk,
  "test_facts".product_sk AS product_sk,
  "merchant_and_product_dims".acquisition_channel AS acquisition_channel,
  SUM("test_facts".amount) AS amount_sum
FROM
  (
    (
      SELECT
        CAST('2023-01-01' AS DATE) AS reporting_date,
        101 AS merchant_sk,
        201 AS product_sk,
        100 AS amount
      UNION ALL
      SELECT
        CAST('2023-01-01' AS DATE) AS reporting_date,
        101 AS merchant_sk,
        203 AS product_sk,
        150 AS amount
      UNION ALL
      SELECT
        CAST('2023-01-02' AS DATE) AS reporting_date,
        102 AS merchant_sk,
        201 AS product_sk,
        200 AS amount
      UNION ALL
      SELECT
        CAST('2023-01-02' AS DATE) AS reporting_date,
        103 AS merchant_sk,
        202 AS product_sk,
        200 AS amount
    )
  ) AS "test_facts"
  LEFT JOIN (
    (
      SELECT
        'M1' AS merchant_id,
        'P1' AS product_id,
        'Organic' AS acquisition_channel
      UNION ALL
      SELECT
        'M1' AS merchant_id,
        'P2' AS product_id,
        'Paid' AS acquisition_channel
      UNION ALL
      SELECT
        'M2' AS merchant_id,
        'P1' AS product_id,
        'Referral' AS acquisition_channel
    )
  ) AS "merchant_and_product_dims" ON "merchant_dims".merchant_id = "merchant_and_product_dims".merchant_id
  AND "product_dims".product_id = "merchant_and_product_dims".product_id
GROUP BY
  1,
  2,
  3,
  4

Expected behavior
Cube should recognize that the join to merchant_and_product_dims depends on merchant_dims and product_dims being joined to test_facts first. Therefore, merchant_dims and product_dims should be included in the FROM clause.

-- Expected SQL
SELECT
  "test_facts".reporting_date,
  "test_facts".merchant_sk AS merchant_sk,
  "test_facts".product_sk AS product_sk,
  "merchant_and_product_dims".acquisition_channel AS acquisition_channel,
  SUM("test_facts".amount) AS amount_sum
FROM
  (
    (
      SELECT
        CAST('2023-01-01' AS DATE) AS reporting_date,
        101 AS merchant_sk,
        201 AS product_sk,
        100 AS amount
      UNION ALL
      SELECT
        CAST('2023-01-01' AS DATE) AS reporting_date,
        101 AS merchant_sk,
        203 AS product_sk,
        150 AS amount
      UNION ALL
      SELECT
        CAST('2023-01-02' AS DATE) AS reporting_date,
        102 AS merchant_sk,
        201 AS product_sk,
        200 AS amount
      UNION ALL
      SELECT
        CAST('2023-01-02' AS DATE) AS reporting_date,
        103 AS merchant_sk,
        202 AS product_sk,
        200 AS amount
    )
  ) AS "test_facts"
  LEFT JOIN (
    SELECT
      101 AS merchant_sk,
      'M1' AS merchant_id
    UNION ALL
    SELECT
      102 AS merchant_sk,
      'M1' AS merchant_id
    UNION ALL
    SELECT
      103 AS merchant_sk,
      'M2' AS merchant_id
  ) AS "merchant_dims" -- This join should be present
  ON "test_facts".merchant_sk = "merchant_dims".merchant_sk
  LEFT JOIN (
    SELECT
      201 AS product_sk,
      'P1' AS product_id
    UNION ALL
    SELECT
      202 AS product_sk,
      'P1' AS product_id
    UNION ALL
    SELECT
      203 AS product_sk,
      'P2' AS product_id
  ) AS "product_dims" -- This join should be present
  ON "test_facts".product_sk = "product_dims".product_sk
  LEFT JOIN (
    (
      SELECT
        'M1' AS merchant_id,
        'P1' AS product_id,
        'Organic' AS acquisition_channel
      UNION ALL
      SELECT
        'M1' AS merchant_id,
        'P2' AS product_id,
        'Paid' AS acquisition_channel
      UNION ALL
      SELECT
        'M2' AS merchant_id,
        'P1' AS product_id,
        'Referral' AS acquisition_channel
    )
  ) AS "merchant_and_product_dims" ON "merchant_dims".merchant_id = "merchant_and_product_dims".merchant_id
  AND "product_dims".product_id = "merchant_and_product_dims".product_id
GROUP BY
  1,
  2,
  3,
  4

Version:
v1.3.6