Unset noise after noise tests

[bmhowe23]

This helps fix issue #609 by unsetting noise models at the end of noise tests. Without this change, the prior test's noise model could be used in subsequent tests, which is undesirable.

(Edited later) The following information is no longer applicable but is left for posterity...

In addition, a new test was added to show another related issue existed if the noise model was used outside of its original scope (as the API permits).

The resolution for that is to convert the noise model to shared_ptr instead of raw pointers. It would've been preferable to use unique_ptr, but Python bindings do not allow unique_ptr's to be used as function arguments, and we need the ability pass them into the C++ engine via functions, so shared_ptr was chosen.

[github-actions]

CUDA Quantum Docs Bot: A preview of the documentation can be found here.

[amccaskey]

Thanks for the fixes here, Ben. I would like to try to find a different approach, as I do not want to make it a requirement that users instantiate a shared_ptr at the language level. They should be free to do so, but we should not enforce it. As for the Python unique_ptr issue, make_kernel() returns a unique_ptr and we pass this instance to other API functions that are bound to Python. In this case I think we just need to be careful about passing by reference vs value in the pybind bindings.

I think there are other approaches that we can take to address the issues we've seen in the runtime. Happy to meet this week to discuss those approaches. One approach may be an update to the noise model destructor to call unset_noise.

[bmhowe23]

Thanks for the feedback. I'm definitely open to different approaches and was indeed questioning if this was the best path forward. Do you think 447b7ce is a valid use case that should be supported? Maybe the answer is 'no', and if so, maybe the smart pointer portion of these changes aren't needed.

[github-actions]

CUDA Quantum Docs Bot: A preview of the documentation can be found here.

[bmhowe23]

After further discussion with @amccaskey, we decided to remove the smart pointer portion of this PR and instead just a) make it clear when the user must call unset_noise(), and b) update our tests to be compliant with that guidance.

[github-actions]

CUDA Quantum Docs Bot: A preview of the documentation can be found here.

[github-actions]

CUDA Quantum Docs Bot: A preview of the documentation can be found here.