Loss progress tests

[Chase-Grajeda]

Implements changes for Jira BF-28: Added tests to inspect loss progress and loss on the progress bar

Changes: test_approximators/test_fit.py

• Added test_loss_progress to monitor the training loss progression in stdout. This is done by checking that the loss reported for each batch is not NaN, and ensuring that the captured epoch summary loss and validation loss match the model's history

test_approximators/conftest.py

• Updated inference_network fixture to use CouplingFlow
• Added inference variables and conditions to approximator fixture
• Removed dataset fixture
• Added simulator fixture to use new NormalSimulator
• Added train_dataset fixture
• Added validation_dataset fixture

simulators/normal_simulator.py

• Added NormalSimulator with sampling for fit tests

simulators/__init__.py

• Added import for NormalSimulator

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests :white_check_mark:

Project coverage is 66.82%. Comparing base (49bbe79) to head (2da5ec7). Report is 10 commits behind head on streamlined-backend.

:exclamation: Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

```diff @@ Coverage Diff @@ ## streamlined-backend #176 +/- ## ======================================================= - Coverage 67.62% 66.82% -0.80% ======================================================= Files 88 92 +4 Lines 1640 1688 +48 ======================================================= + Hits 1109 1128 +19 - Misses 531 560 +29 ```

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.