Automatically test plxpr capture integration with assert_valid

[albi3ro]

Context:

Following on from #5523, we need a good way of ensuring operations will integrate well with program capture.

Description of the Change:

This PR adds a check to the assert_valid function to ensure capture is working as expected.

It then starts to fix up some operations that were failing to be captured correctly.

Benefits:

More robust integration with pl capture.

Possible Drawbacks:

I'm realizing a lot more operators fail to be immediately captured than I initially assumed 😢

Related GitHub Issues:

[sc-63310]

[dwierichs]

LGTM, just a few tiny questions. Thanks for the patch @albi3ro!
I'm wondering: Can we have a systematic way to make sure that assert_valid is called on all PL objects it can be applied to? 🤔
Similar to the logic in the templates testing file, where PL is scraped for Operators (and SymbolicOps?) and assert_valid keeps track of which classes it saw? 🤔

[dwierichs]

Thanks for the explainers! 🙏

[albi3ro]

LGTM, just a few tiny questions. Thanks for the patch @albi3ro! I'm wondering: Can we have a systematic way to make sure that assert_valid is called on all PL objects it can be applied to? 🤔 Similar to the logic in the templates testing file, where PL is scraped for Operators (and SymbolicOps?) and assert_valid keeps track of which classes it saw? 🤔

So by default we do automatically validate all operators except for templates. See this file for the automatic collection: https://github.com/PennyLaneAI/pennylane/blob/master/tests/ops/functions/conftest.py

We opted templates out of this process due to the complexity of automatically generating a valid instance, but most templates have an assert_valid test associated with them.

[dwierichs]

Ah, brilliant! So the only thing one could add is a cache that memorizes tested templates (from the template test files) and checks for untested templates (as per assert_valid). I can see this being very shaky though 🙃

[dwierichs]

Two new comments I had.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 99.67%. Comparing base (c14ce7b) to head (3e5ef90).
Report is 256 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #5686      +/-   ##
==========================================
- Coverage   99.67%   99.67%   -0.01%     
==========================================
  Files         414      414              
  Lines       39316    39167     -149     
==========================================
- Hits        39188    39038     -150     
- Misses        128      129       +1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.