Vectorize dispatch for shape operations

[ricardoV94]

It also fixes a bug when Blockwising a Reshape Op, as the perform method had a non-optional (useless) params kwarg

[codecov-commenter]

Codecov Report

Merging #454 (bd72bc7) into main (36df379) will increase coverage by 0.01%.
Report is 4 commits behind head on main.
The diff coverage is 100.00%.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #454      +/-   ##
==========================================
+ Coverage   80.66%   80.67%   +0.01%     
==========================================
  Files         160      160              
  Lines       46025    46060      +35     
  Branches    11266    11277      +11     
==========================================
+ Hits        37124    37159      +35     
  Misses       6668     6668              
  Partials     2233     2233              

Files	Coverage Δ
pytensor/graph/replace.py	86.45% <100.00%> (+0.28%)	⬆️
pytensor/tensor/blockwise.py	79.05% <100.00%> (-0.22%)	⬇️
pytensor/tensor/elemwise.py	88.59% <100.00%> (-0.02%)	⬇️
pytensor/tensor/random/op.py	96.25% <ø> (ø)
pytensor/tensor/shape.py	92.91% <100.00%> (+0.52%)	⬆️
pytensor/tensor/type.py	94.52% <100.00%> (+0.04%)	⬆️

[jessegrabowski]

Just hit this bug, so I'm suddenly motivated to review : )

Looks great. Small nitpicks if you want them: There's some code duplication for error handling between the dispatchers for specify shape and reshape that could be refactored. Same with the test cases -- could probably just use a mark.parameterize decorator to pass in which Op to test, and make the actual test cases shared?

None of this blocks tho

[ricardoV94]

The two are ever so slightly different because reshape takes a vector and SpecifyShape takes individual dim lengths. I think it's better to jut be a bit repetitive here for the sake of more clear code.