Add gammainc(c) derivatives

[ricardoV94]

This PR adds a Python implementation of derivatives of the gammainc and gamaincc. Closes #74.

The algorithms were adapted from the STAN library.

[codecov]

Codecov Report

Merging #513 (13d3984) into main (0ae63d1) will increase coverage by 0.01%.
The diff coverage is 82.79%.

❗ Current head 13d3984 differs from pull request most recent head 924e8c4. Consider uploading reports for the commit 924e8c4 to get more accurate results

@@            Coverage Diff             @@
##             main     #513      +/-   ##
==========================================
+ Coverage   76.89%   76.91%   +0.01%     
==========================================
  Files         149      149              
  Lines       46602    46689      +87     
  Branches    10230    10242      +12     
==========================================
+ Hits        35833    35909      +76     
- Misses       8190     8196       +6     
- Partials     2579     2584       +5     

Impacted Files	Coverage Δ
aesara/scalar/math.py	84.52% <82.79%> (+0.43%)	⬆️

[twiecki]

Is speed a concern here with a pure-Python implementation?

[ricardoV94]

Is speed a concern here with a pure-Python implementation?

Definitely. We have to fix #512 before deciding on how to proceed in #514 (pure Aesara function or this Op with specialized C/Numba dispatch).

If it's the latter we still want to keep the Python code for fast compilation mode

[brandonwillard]

Do you want to create a follow-up issue for the non-Python implementation and merge this in the meantime?

[ricardoV94]

Do you want to create a follow-up issue for the non-Python implementation and merge this in the meantime?

There is already a follow up issue here: #514

Yes, I think we can merge this already

[brandonwillard]

The last commit was a merge commit; I've rebased instead.