re-add support for tf.function with abstract shape

[timmysilv]

Context:
While working on adding poetry support, this qcut test failed in a familiar way - it tried to check batching with an abstract TensorShape (of None)! Sub-context: it hasn't been run for a very long time because it needs networkx<3 and poetry is making me use that, unlike today's CI runs. I thought I had fixed this in #4911 mostly because tests stopped failing. However, they (eg. the test that I am changing here) stopped failing because we were no longer running the batch-size evaluation. It's a niche corner case (see that PR description for more details), but it used to pass and it now fails on master.

Description of the Change:

• Re-add the previous functionality to Operator._check_batching() where it skips the check if the operator params are abstract TF tensors with a shape of None in its signature within a tf.function
• Removed the MyRX from the test because it only made sense when I added it (in other words, before I re-defined Operator.ndim_params in the linked PR), but am now calling op.batch_size to ensure the loading of the property

Benefits:
Tests that used to pass are passing again, specifically if you use tf.function with input_signature=(tf.TensorSpec(shape=None))

Possible Drawbacks:
Additional try-except logic (re-)added to Operator._check_batching() which is called frequently

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (4e2e344) 99.68% compared to head (43234dc) 99.68%.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #5089      +/-   ##
==========================================
- Coverage   99.68%   99.68%   -0.01%     
==========================================
  Files         392      392              
  Lines       35844    35578     -266     
==========================================
- Hits        35732    35465     -267     
- Misses        112      113       +1     

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

[mudit2812]

I recall we made batch checking lazy for tapes. Curious to see how much of a difference this PR actually makes with that change in mind. Do you have any light benchmarks?

[timmysilv]

not personally, I do believe @albi3ro mentioned the performance improvement so she might have some numbers handy. Either way, this change is now concerned with correctness because I broke something that used to work, so I'd like to restore functionality

[timmysilv]

[sc-55335]