Make Wofost72 differentiable and vectorized

[SCiarella]

This PR closes #49.

Noticeably, Wofost72 uses kiosk variables such as EVS that come from the soil model. So in order to make wofost72 consistently differentiable and vectorized, I had to implement a differentiable soil model soil/classic_waterbalance.py which is inspired by the corresponding model from pcse.

This PR also adds a lot of tests, making issue #81 more relevant than ever.

[SCiarella]

In order to reduce the execution time, I have implemented the following modifications:

1. ComputeConfig now stores a flag _check_carbon_balance that is used by wofost72 to skip some of the expensive checks that slow down the integration [Removed]

2. assimilation now has a cache to store the constants and share them between different submodules. More importantly, the 3-point quadrature methods have been refactored in order to be vectorized.

3. phenology uses torch.where instead of if statements to handle mixed states

4. removed the pcse.decorators that are redundant with Tensor

[SCiarella]

In order to make the tests more manageable (the test action takes >2h if unchecked), I have introduced two pytest parser addoptions:

1. --full_wofost72_test if used, runs the gradient test for all the parameter combinations used by wofost72. By default, we now only test the gradients of the parameters that are directly manipulated in wofost72.
2. --fast when specified, it: (a) makes pytest use test files 0 to 5 instead of all the 45 available, and (b) skips GPU and tensor-input in all the gradient tests.

@SarahAlidoost, we could think about using --fast by default for the GitHub actions to significantly save time, and only run the tests without --fast once before PR merge. This is probably the only real solution to #81. What do you think?

[sonarqubecloud]

Quality Gate passed

Issues
15 New issues
0 Accepted issues

Measures
0 Security Hotspots
95.5% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud