Fix str.replace(..., n=0) replacing every occurrence instead of none

[Kropiunig]

Description

StringAccessor.replace documents n as "Number of replacements to make from start. Use -1 to replace all." The regex code path (the default, regex=True) forwards count=max(n, 0) to re.Pattern.sub, and re.sub treats count=0 as replace every occurrence. So n=0 collapses onto n=-1 and replaces everything instead of nothing — silently, with no error.

The literal path (regex=False) calls str.replace(pat, repl, n), which honours n=0 correctly. The same call therefore gives two different answers depending on regex:

>>> import xarray as xr
>>> da = xr.DataArray(["fooBAD__barBAD"], dims="x")

>>> da.str.replace("BAD", "", n=0).item()          # regex=True (default)
'foo__bar'                                          # expected 'fooBAD__barBAD'

>>> da.str.replace("BAD", "", n=0, regex=False).item()
'fooBAD__barBAD'                                    # correct

This is most damaging where n is array-like, which is a documented feature ("If pat, repl, or n is array-like, they are broadcast against the array and applied elementwise"). A single 0 in the broadcast n silently becomes "replace all" for that element:

>>> n = xr.DataArray([0, 1, -1], dims="y")
>>> da.str.replace("BAD[_]*", "", n=n).values
array([['foobar', 'foobarBAD', 'foobar']], dtype='<U9')
#        ^^^^^^ should be 'fooBAD__barBAD'

Two further references agree that n=0 means "no replacements": pandas.Series.str.replace(..., n=0) is a no-op for both regex=True and regex=False, and xarray's own str.split(..., maxsplit=0) already performs zero splits.

The fix special-cases n == 0 in the regex branch. Negative n still maps to "replace all" exactly as before, so no other behaviour changes.

Added test_replace_n_zero, which covers the regex path, the literal path and the broadcast case. It fails on main (2 failed / 12 passed for -k replace) and passes with this change (14 passed); the full xarray/tests/test_accessor_str.py module is 228 passed, 1 skipped.

Checklist

• Tests added
• User visible changes (including notable bug fixes) are documented in whats-new.rst

AI Disclosure

• This PR contains AI-generated content.
  • I have tested any AI-generated content in my PR.
  • I take responsibility for any AI-generated content in my PR.
    Tools: Claude Code. Prompt: audit the xarray source for a provably-correct, unclaimed bug with a user-visible symptom on the default code path, then produce a minimal fix plus a regression test that fails before and passes after.

[welcome]

Thank you for opening this pull request! It may take us a few days to respond here, so thank you for being patient.
If you have questions, some answers may be found in our contributing guidelines.