[Return-types #5.3] Param-shift shot vector case

[antalszava]

Context:
Follows the previous PRs on Return-types to allow the param_shift gradient transform to process results when the gradient tapes are executed on a device that has a shot vector defined.

Description of the Change:

• Introduced the shots parameter to the param_shift, expval_param_shift and var_param_shift functions including the new return type versions of these functions and auxiliary functions that may require information about shots;
• Updated the logic of param_shift to handle executing gradient tapes on a device with a shot vector;
• Ported over the majority of tests from tests/returntypes/test_parameter_shift_new.py (Hamiltonian tests being left as TODO).

Benefits:
Most cases work for execution using shot vectors.

Possible Drawbacks:
The logic that attempts to determine if the user didn't pass the shots argument, but execution happened with a shot vector device. To do this, ValueErrors raised are being caught - the user sees a message related to setting the shots argument along the original error message.

Related GitHub Issues:

[sc-27498]

[rmoyard]

Great job @antalszava, it is looking good 💯 I have some comments and questions. It might be worth to create a smalll separate test class with all the cases we know where you parametrize the number of parameters, number of measurements and different shot vectors like we did for backprop previously.

[timmysilv]

my above comments are all pretty small, this PR looks great! updating that test file was a true feat 😸

[antalszava]

@timmysilv @rmoyard thanks for the great comments! 🎉

I think I've addressed them all, some with questions - keen to get further thoughts. 👍 Let me know what should definitely change still in this PR.

[timmysilv]

🥳

[rmoyard]

Great job @antalszava 🥇 We might be missing one test case. If it is not the case I will be happy to approve the PR!

[rmoyard]

Thanks it looks great 💯