rolling_mean_pad Performance Improvements

[TimothyWillard]

Replaced custom rolling mean implementation using np.pad and np.mean with implementation based on scipy.ndimage.convolve1d. This implementation also maintains the window shrinking behavior of the current implementation by manually computing the last row.

Addresses GH-248, but there aren't any unit tests for rolling_mean_pad right now AFAICT so I would feel much more confident merging this in after GH-247 which adds unit tests for this function (and will probably generate some merge conflicts).

[jcblemai]

Approved, thank you :)

[TimothyWillard]

I've run these changes through the test suite written in GH-247 and now feel confident in merging this into main.

(venv) twillard@epid-iss-mbp ~/D/G/H/f/f/gempyor_pkg (performance/GH-248-rolling_mean_pad-speed)> pytest tests/utils/test_rolling_mean_pad.py
=============================================================================================== test session starts ================================================================================================
platform darwin -- Python 3.11.9, pytest-8.2.2, pluggy-1.5.0
rootdir: /Users/twillard/Desktop/GitHub/HopkinsIDD/flepiMoP/flepimop/gempyor_pkg
configfile: pyproject.toml
collected 50 items

tests/utils/test_rolling_mean_pad.py ..................................................                                                                                                                      [100%]

================================================================================================ 50 passed in 0.78s ================================================================================================