Feature/mask NaNs in training loss function

[sahahner]

Variables with missing values that are imputed by the imputer should not be considered in the loss.

Solves issue 271

Describe the solution
Pass the product of the imputer NaN mask(s) to the loss in the first forward pass and multiply the contributions to the loss of imputed grid points by zero.
Also, apply the remapper to the mask to remap the NaN mask in case the remapper is used.

The NaN masks are prepared in the imputer. The remapper contains a new function to remap the NaN mask.
These changes are part of anemoi-model, PR #56

Attention
This changes the default behaviour when using variables that contain NaN values that are imputed.

[FussyDuck]

All committers have signed the CLA.

[HCookie]

The loss function code is being altered in #70 to enable flexible configuration of loss functions.
This PR will require some changes to be compatible.

Changes

• masks will need to be added to ‘mae’, and ‘logcosh’
• Initialisation of loss functions needs to be pulled from the other PR
• Setting of the mask will require a ‘hasattr’ check to determine if the loss function exposes that functionality.

[sahahner]

This has to be merged after #137

[jakob-schloer]

After discussing with @sahahner and a quick test with an ocean dataset, I approve the PR.

[jakob-schloer]

Approve after testing.