Spectrum Viewer: Prevent QThread from accessing Presenter data during runtime

[MikeSullivan7]

Issue Closes #2715

Description

The PR is to prevent the QThread in the Spectrum Viewer from changing data which may be accessed by different processes, while the QThread is running as this may lead to unreliable results due to a race condition. To achieve this, the function in the Worker Thread no longer directly accesses the attributes in the presenter, but instead takes them as arguments and outputs the altered attributes as thread.result(). The attributes in the presenter space are then changed during the thread_cleanup instead of inside the thread itself.

Also to make sure that nothing could be changed in the running of the thread, a Lock() is used inside the thread function.

Developer Testing

• I have verified unit tests pass locally: python -m pytest -vs

Acceptance Criteria and Reviewer Testing

• Unit tests pass locally: python -m pytest -vs
• Load data in the spectrum viewer and open multiple ROIs
• Move the position and sizes of the ROIs in any combination of ways you can think of and as fast as you feasibly can.
• ensure that the spectrums update as expected and that no errors are raised.

[coveralls]

coverage: 70.37% (+0.08%) from 70.295%
when pulling faf4a94 on 2715_spectrum_thread_mutable
into 3b7fb16 on main.

[MikeSullivan7]

The kwargs for the TaskWorkerThread has now been changed such that the function receives a copy of the input data which it then modifies, instead of modifying the attributes of the presenter directly. The presenter attributes are then changed in the thread cleanup.
This can be checked via e.g.

print(f"{(self.roi_to_process_queue is self.thread.result[0])=}")
print(f"{(self.image_nan_mask_dict is self.thread.result[1])=}")
print(f"{(self.view.spectrum_widget.spectrum_data_dict["roi"] is self.thread.result[2]["roi"])=}")

In the thread_cleanup() method

The only thing which is not local in run_spectrum_calculation now is the use of self.model.get_spectrum() but as this is purely a getter, this will not modify any non-local data from inside the worker thread.

[samtygier-stfc]

Would it work to move all the chunk logic into handle_roi_moved() and for i in range(len(spectrum)): into thread_cleanup()? it is only really the self.model.get_spectrum() call that is slow an needs to be in the thread. I think that would save from needing the copies, because its only some simple values being passed in, and the spectrum array being passed back.

[ashleymeigh2]

Looks good