Fix DataFrame.applymap deprecation on pandas >= 2.1 (keeps pandas < 2.1 working)

[goosfrabba]

Fixes #258.

On pandas >= 2.1, DataFrame.applymap is deprecated in favour of DataFrame.map, so parallel_applymap currently raises FutureWarning: DataFrame.applymap has been deprecated. Use DataFrame.map instead (from pandarallel/data_types/dataframe.py).

This makes the elementwise apply version-aware: it uses DataFrame.map when the installed pandas provides it (>= 2.1) and falls back to DataFrame.applymap otherwise. That resolves the deprecation without dropping support for pandas < 2.1 and without raising the minimum Python version, addressing the concern raised on #259 (a hard applymap -> map swap).

• pandarallel/data_types/dataframe.py: ApplyMap.work picks map/applymap by capability.
• tests/test_pandarallel.py: the reference in test_dataframe_applymap uses the same capability check so it stays warning-free across pandas versions.

A sensible follow-up is extending the CI matrix to pandas 2.1/2.2 and Python 3.11-3.13 so this class of regression is caught automatically.

[codecov]

Codecov Report

❌ Patch coverage is 0% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 79.55%. Comparing base (261a652) to head (f4e59d5).

Files with missing lines	Patch %	Lines
pandarallel/data_types/dataframe.py	0.00%	2 Missing ⚠️

❗ There is a different number of reports uploaded between BASE (261a652) and HEAD (f4e59d5). Click for more details.

HEAD has 58 uploads less than BASE

Flag	BASE (261a652)	HEAD (f4e59d5)
	63	5

Additional details and impacted files

@@             Coverage Diff             @@
##           master     #280       +/-   ##
===========================================
- Coverage   91.28%   79.55%   -11.73%     
===========================================
  Files          12       12               
  Lines         585      587        +2     
===========================================
- Hits          534      467       -67     
- Misses         51      120       +69     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[goosfrabba]

Heads up on the red CI: the macOS failures are environment rot in the old-pandas matrix, not related to this patch — pandas 1.x wheels hitting the numpy 2.x ABI break (numpy.dtype size changed), and pandas source builds failing on newer setuptools (No module named 'pkg_resources'). Pinning numpy<2 for the pandas<2 jobs (and setuptools<81 for the 3.9 source builds) should green them. Happy to add that to this PR or a separate one if useful.

[LalwaniPalash]

Some test evidence in support of this, from running the suite on pandas 3.0.3 / Python 3.14.6.

Worth flagging that on pandas 3.0 this is no longer a deprecation — DataFrame.applymap has
been removed outright:

AttributeError: 'DataFrame' object has no attribute 'applymap'

So parallel_applymap currently raises for every caller on pandas 3.x, and
test_dataframe_applymap accounts for 16 of the 24 failures in the suite. The hasattr(data, "map")
guard here handles that correctly — DataFrame.map exists from pandas 2.1 onward, so the fallback
still covers older versions.

The other 8 failures are unrelated (grouping columns are no longer passed to apply on pandas 3.0);
I've opened #282 for those. Combining the two branches locally:

	failed	passed
master	24	193
master + #282	16	201
master + #280	8	209
master + both	0	217

One caveat on CI: the workflow matrix targets ubuntu-20.04, which GitHub has retired, so those
jobs are cancelled before they run — that's why this PR shows red rather than anything to do with
the change itself. The matrix also tops out at pandas 2.0, so it wouldn't exercise this fix either
way.