Refactor Numba dispatching and interfaces

[brandonwillard]

This PR clarifies some of the dispatch interfaces for the Numba transpilation code, by—for instance—making it possible to customize Numba signature construction based on Aesara Types. It also introduces some necessary updates for the use of more recent Numba versions.

[codecov]

Codecov Report

Merging #1317 (43304c8) into main (14c394d) will increase coverage by 0.12%.
The diff coverage is 93.31%.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1317      +/-   ##
==========================================
+ Coverage   74.10%   74.22%   +0.12%     
==========================================
  Files         174      174              
  Lines       48673    48754      +81     
  Branches    10373    10367       -6     
==========================================
+ Hits        36067    36190     +123     
+ Misses      10315    10275      -40     
+ Partials     2291     2289       -2     

Impacted Files	Coverage Δ
aesara/link/jax/dispatch/basic.py	83.87% <ø> (ø)
aesara/link/jax/dispatch/scalar.py	97.40% <ø> (ø)
aesara/misc/check_blas.py	0.00% <0.00%> (ø)
aesara/misc/elemwise_time_test.py	0.00% <0.00%> (ø)
aesara/misc/latence_gpu_transfert.py	0.00% <0.00%> (ø)
aesara/compile/function/types.py	79.16% <75.00%> (+0.16%)	⬆️
aesara/link/c/cmodule.py	52.04% <77.77%> (ø)
aesara/tensor/slinalg.py	92.59% <81.15%> (+8.18%)	⬆️
aesara/tensor/elemwise.py	88.07% <90.54%> (-0.52%)	⬇️
aesara/graph/rewriting/basic.py	65.03% <92.30%> (ø)
... and 34 more