docs: document regular expression function arguments

[stantheman0128]

Which issue does this PR close?

Related to #1463.

Rationale for this change

regexp_instr documents every parameter it takes in an Args: section. The four
other regexp_* functions document none. Someone reading the generated API docs to
find out what flags accepts, or where start counts from, gets an answer for one
function in the family and nothing for the rest.

This is the same kind of small focused pass as #1527, applied to the regular
expression family.

What changes are included in this PR?

Adds an Args: section to four functions in python/datafusion/functions/__init__.py:

• regexp_like
• regexp_match
• regexp_replace
• regexp_count

Where a parameter also exists on regexp_instr (regex, start, flags), the
wording follows what regexp_instr already says, so the family reads consistently.
The flags entry on regexp_replace also records the g behavior that the prose
above it and its own example already show.

Nothing else in the file changes. Signatures, type hints, runtime code, and existing
examples are untouched.

Are there any user-facing changes?

Yes, documentation only. The generated API docs now list argument descriptions for
these four functions. Runtime behavior is unchanged.

Validation

Pre-commit, scoped to the changed file:

$ pre-commit run --files python/datafusion/functions/__init__.py
Lint GitHub Actions workflow files...................(no files to check)Skipped
ruff (legacy alias)......................................................Passed
ruff format..............................................................Passed
Rust fmt.............................................(no files to check)Skipped
Rust clippy..........................................(no files to check)Skipped
codespell................................................................Passed
uv-lock..............................................(no files to check)Skipped

Ruff on its own, pinned to the version in .pre-commit-config.yaml:

$ ruff@0.15.1 check --config pyproject.toml python/datafusion/functions/__init__.py
All checks passed!
$ ruff@0.15.1 format --check --config pyproject.toml python/datafusion/functions/__init__.py
1 file already formatted

The repository enables --doctest-modules over python/datafusion, so the examples
in these docstrings run as tests. Against a locally built extension:

$ pytest python/datafusion/functions/__init__.py -q -k regexp
5 passed, 231 deselected, 2 warnings in 2.39s

$ pytest python/datafusion/functions/__init__.py -q
236 passed, 2 warnings in 2.06s

Whitespace:

$ git diff --check

LLM-generated code disclosure

These docstring additions were prepared with the assistance of Claude and reviewed
against each function signature before submission.

[ntjohnson1]

Nit about not repeating information already captured in doc string since we can test the validity of doc strings but can't test the free text stays up to date. Applies to all the optional call outs.

[ntjohnson1]

Suggested change

	flags: Optional regular expression flags to control regex behavior.
	flags: Flags to control regex behavior.

NIT: Optional is implied by the = None

[stantheman0128]

Applied to all five: the three plain flags lines, the flags line on regexp_replace that also documents g, and start on regexp_count.

Worth flagging, since your reasoning has a consequence I did not expect. regexp_instr documents its own parameters the same way I originally did (flags: Optional regular expression flags..., start: Optional position...), and I copied that wording precisely so the family would read consistently. After this change the four read differently from the fifth, in the opposite direction.

I can include regexp_instr here so the family lands consistent, or leave it for a separate pass if you would rather keep this PR to the four. Your call.