Add helper to build hessian vector product

[ricardoV94]

Description

This can be quite useful for optimization, and it's perhaps not trivial for users how to build the expression (it surely wasn't for me)

Actually this was described in the docs, I still think a helper may be easier / more discoverable: https://pytensor.readthedocs.io/en/latest/tutorial/gradients.html#hessian-times-a-vector

[ricardoV94]

Failing tests are probably due to changes in new scipy that was released 2 days ago

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 81.20%. Comparing base (f4e249d) to head (ce484de).
Report is 117 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #858   +/-   ##
=======================================
  Coverage   81.20%   81.20%           
=======================================
  Files         170      170           
  Lines       46992    47001    +9     
  Branches    11510    11512    +2     
=======================================
+ Hits        38160    38169    +9     
  Misses       6621     6621           
  Partials     2211     2211           

Files with missing lines	Coverage Δ
pytensor/gradient.py	77.64% <100.00%> (+0.27%)	⬆️

[aseyboldt]

I don't think there is any reason to require wrt and p to be 1d. As long as wrt and p have the same shape(s) everything should be fine.

Maybe we can add a test for that and remove the "1-dimensional" from the docs?
Otherwise this look good.

[ricardoV94]

I don't think there is any reason to require wrt and p to be 1d. As long as wrt and p have the same shape(s) everything should be fine.

Maybe we can add a test for that and remove the "1-dimensional" from the docs? Otherwise this look good.

Sorry @aseyboldt I missed your comment before pressing the green button! Opening an issue