Differentiable Model Fixtures

[leoglonz]

Problem

In a previous issue we bolstered some of the fixtures in dMG. Once all the MHPI differentiable models (i.e., dHBV1.0, dHBV1.1p, dHBV2.0) have been validated on the framework to behave identical to hydrodl, then we can create specific fixtures to warn us whenever the behavior of these models in train/test deviates from norm.

Proposed Solution

Basically, the solution is above. We need to have fixtures specific to all published differentiable models that will perform a validation on models train/test with very small datasets (this has already been done for dHBV in tests/test_train_regression.py).

Additional context