opt: normalize sequential % in regexes to single %

[mgartner]

There are several optimization the execution engine can perform when evaluating LIKE patterns with %, which matches zero or any number of any character. For example, see #80073.

Sequential %s can prevent these optimizations, but they are semantically equivalent to a single %, so the optimizer should normalize them to a single %.

For example:

-- The following expressions are equivalent:
col LIKE '%%abc%%%%'
col LIKE '%abc%'

Jira issue: CRDB-15814

[ADustyOldMuffin]

I'd be interested in taking this on.