Deprecation

[charlie-becker]

1. Added a deprecated module with the old models, losses, and callbacks.
2. Modified losses.py to remove old losses and add new ones
3. Modified callbacks.py to remove old ones and add new ones

[djgagne]

The CI had the following error

ImportError while importing test module '/home/runner/work/miles-guess/miles-guess/mlguess/tests/test_models.py'.
Hint: make sure your test modules/packages have valid Python names.
Traceback:
../../../micromamba/envs/guess/lib/python3.10/importlib/__init__.py:126: in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
mlguess/tests/test_models.py:9: in <module>
    from mlguess.keras.models import BaseRegressor as RegressorDNN
mlguess/keras/models.py:13: in <module>
    from mlguess.keras.losses import EvidentialRegressionLoss, EvidentialRegressionCoupledLoss, gaussian_nll
E   ImportError: cannot import name 'EvidentialRegressionLoss' from 'mlguess.keras.losses' (/home/runner/work/miles-guess/miles-guess/mlguess/keras/losses.py)

[djgagne]

@charlie-becker it looks like you need to update your tests to use the new losses and remove or refactor the imports for the old losses.

[charlie-becker]

Model classes have been fully refactored. I have removed the tests for the non-evidential models (as they have been moved to the deprecated module).

.predict() functions in both of the new model classes (classification and regression) now include a return_uncertainties= argument (bool).

This should thoroughly documented in the README which I can accomplish in another PR.

[djgagne]

Looks good to me. Merging.