Align `regexp_like`, `~`, and `~*` implementations

[pepijnve]

Is your feature request related to a problem or challenge?

The implementation of regexp_like is less optimised compared to the ~ and ~* operators. In particular, the operators have optimised paths for array ~ scalar while regexp_like will expand scalars to arrays. The consequence is that the function will always use the regexp_is_match kernel while the operators can make use of regexp_is_match_scalar. This in turn results in regex_like(column, 'pattern') having different performance characteristics compared to column ~ pattern which can be a pitfall.

Describe the solution you'd like

Try to align the two implementations as well as possible so there is no practical difference between the two.

Describe alternatives you've considered

No response

Additional context

No response

[pepijnve]

The PR I made for this implements a small improvement for common cases by rewriting regexp_like expressions into operator expressions.

An alternative would be to extract the regex support code from physical_expr::expressions::binary to a separate module regex sub project so that it can be used from functions as well. That's probably the way to go to long term to deduplicate the code entirely, but that would be a much bigger change.