Closes #4650

This ended up being a bigger change than I intended. Some notes:

• I followed the groupings in the linked issue per format.
• I removed the regex-based approach for matching formats for simple string-matching. I didn't quite see an easy way to reuse it, as underscores can appear in any position in a literal. Makes it so we get the "fixing" functionality for semi-free.
  • I did reuse the regex's in the tests.
• NumericUnderscores is another extension, now the code action can add multiple of these.