Implement multinomial and inverse gamma in numba backend

[ricardoV94]

Also changes the python implementation to use numpy instead of scipy. The overhead doesn't make sense for this trivial RV.

Long term we should replace these by OpFromGraph: #1221

---

📚 Documentation preview 📚: https://pytensor--1223.org.readthedocs.build/en/1223/

[codecov]

Codecov Report

Attention: Patch coverage is 50.00000% with 14 lines in your changes missing coverage. Please review.

Project coverage is 82.01%. Comparing base (e25e8a2) to head (7fdfe64).
Report is 3 commits behind head on main.

Files with missing lines	Patch %	Lines
pytensor/link/numba/dispatch/random.py	44.00%	14 Missing ⚠️

❌ Your patch status has failed because the patch coverage (50.00%) is below the target coverage (100.00%). You can increase the patch coverage or adjust the target coverage.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1223      +/-   ##
==========================================
- Coverage   82.03%   82.01%   -0.02%     
==========================================
  Files         188      188              
  Lines       48535    48559      +24     
  Branches     8676     8679       +3     
==========================================
+ Hits        39814    39824      +10     
- Misses       6558     6572      +14     
  Partials     2163     2163              

Files with missing lines	Coverage Δ
pytensor/tensor/random/basic.py	98.84% <100.00%> (ø)
pytensor/link/numba/dispatch/random.py	55.80% <44.00%> (-1.40%)	⬇️