This PR adds a few things at once, sorry.

Breaking

• Changed the 2nd param to be a boolean, toggling long-mode directly
  Of course this can be reverted, and I'm open to it, but don't foresee other options :)

Features

• Multiple formats:

  • ESM (via module entry)
  • CommonJS (via main entry – unchanged)
  • UMD (via unpkg entry)

• Made available on unpkg.com 🎉

Chores

• Adds minimal benchmark suite
• Attaches a light build tool (one of my own) to auto-build ^those formats
• Includes prettier directly – was relying on global installation
• Updates lint script

Here's a before-after of both the module size & the performance:

Running Node v10.13.0, no warmups

Before:

Filename                 Filesize  (gzip)
dist/index.js             1.41 kB  643 B
dist/index.mjs            1.41 kB  633 B
dist/index.min.js         1.56 kB  701 B


# string (short) -> number
  ms   x 11,828,412 ops/sec ±0.46% (92 runs sampled)

# string (long) -> number
  ms   x 5,803,783 ops/sec ±1.78% (96 runs sampled)

# number -> string (short)
  ms   x 76,444,334 ops/sec ±0.89% (96 runs sampled)

# number -> string (long)
  ms   x 33,576,081 ops/sec ±0.69% (93 runs sampled)

# Negative :: string (short) -> number
  ms   x 6,710,932 ops/sec ±1.41% (96 runs sampled)

# Negative :: string (long) -> number
  ms   x 5,944,477 ops/sec ±0.51% (95 runs sampled)

# Negative :: number -> string (short)
  ms   x 80,648,073 ops/sec ±0.22% (96 runs sampled)

# Negative :: number -> string (long)
  ms   x 32,387,825 ops/sec ±0.32% (98 runs sampled)

After

Filename                 Filesize  (gzip)
dist/index.js              948 B   521 B
dist/index.mjs             947 B   521 B
dist/index.min.js         1.10 kB  585 B


# string (short) -> number
  ms   x 13,121,298 ops/sec ±0.89% (93 runs sampled)

# string (long) -> number
  ms   x 6,087,008 ops/sec ±1.90% (93 runs sampled)

# number -> string (short)
  ms   x 77,564,203 ops/sec ±0.60% (91 runs sampled)

# number -> string (long)
  ms   x 46,140,236 ops/sec ±0.29% (93 runs sampled)

# Negative :: string (short) -> number
  ms   x 7,159,583 ops/sec ±0.30% (96 runs sampled)

# Negative :: string (long) -> number
  ms   x 6,152,167 ops/sec ±0.67% (94 runs sampled)

# Negative :: number -> string (short)
  ms   x 78,520,060 ops/sec ±0.64% (96 runs sampled)

# Negative :: number -> string (long)
  ms   x 45,736,035 ops/sec ±0.36% (95 runs sampled)

I have a few other versions locally, ranging from 569B to 487B, but they had different performance trade-offs. This 521B version feels like the best balance IMO.