Specialized / Pre-compiled / Prepared ScalarUDFs

### Is your feature request related to a problem or challenge?

Currently, scalar UDF functions can not be "specialized"

```select
SELECT * FROM t where my_matcher(t.column, '[a-z].*');
```

What happens is that the regexp string `'[a-z].*'` gets passed as a literal expression to each invocation of the function, then you have to compile the regex again and again for every batch you process through the UDF.

Since it is expensive to compile these RegExps, it would be nice if there was something that could compile the RegExp once per plan rather than once per batch

### Describe the solution you'd like

_No response_

### Describe alternatives you've considered

See comments below

### Additional context

Suggested by @thinkharderdev on https://github.com/apache/arrow-datafusion/issues/8045#issuecomment-1793267211

Start with the scalar UDF invocation described in the issue and review the alternatives in the comment thread, including the linked issue discussion. Trace how the query plan handles the literal regular-expression argument and identify the relevant planning and execution entry points. Done means a constant pattern can be prepared once per plan rather than recompiled for every batch, with coverage for the example query.

Written by the indexing model from the issue text.