Pytest framework implementation with Task Runner GitHub Workflow

[noopurintel]

This PR aims to add the pytest framework based test cases in OpenFL repository.

Changes done:

Added end_to_end folder inside openfl/tests. Please go through README.md file for complete structure and related details.
Added task_runner_e2e.yml workflow file inside .github/workflows - it will run various models for Task Runner API approach.
For now, the workflow is expected to run every night with below matrix combination:

1. model names - [torch_cnn_mnist, keras_cnn_mnist] # Other models have bugs associated with them
2. python version - [3.8, 3.9, 3.10]

Successful test run (11th Nov 07:20 pm IST) - https://github.com/securefederatedai/openfl/actions/runs/11779770937

Final (12th Nov 04:35 pm IST) - https://github.com/securefederatedai/openfl/actions/runs/11795815852

[MasterSkepticista]

Suggest naming it as tests/end_to_end, and instances of e2e as end_to_end.

[noopurintel]

@MasterSkepticista - this name openfl_e2e was decided during one of the discussions.
But, the name end_to_end would be easier to comprehend as you have suggested, let me check and get back on this.

[theakshaypant]

LGTM.
Just a few general questions.

Also how do we feel about having the model accuracy to be logged along with test success (or failure) in the workflow? (Can be easily picked up from the aggregator logs.)

[MasterSkepticista]

I have not reviewed the actual tests yet.

[teoparvanov]

This is awesome, thank you @noopurintel ! Very nice abstractions around ModelOwner, Collaborator and Aggregator, which contribute to a great experience both reading and writing the actual tests.

[noopurintel]

@theakshaypant - Fyi, we are planning to add accuracy and other details after every test run in subsequent PRs.

[teoparvanov]

A final comment on method naming...

[MasterSkepticista]

Minor changes.

[MasterSkepticista]

Approving with reservations - I am personally not confident about the scope of what this test framework will achieve for "Taskrunner" interface, and how it helps beyond existing tests. I may be wrong about this, so I agree with merging this PR.

I do see the value of this framework for Workflow API though, to be clear.

Thanks for the additions @noopurintel and @payalcha.