Bridging code to configure which GWP is being used

[dorchard]

• Some variables used in GWPs that are not used in AD99, so work out appropriate data is available at the part of GWP.
• Glue code to convert MiMA data into the correct tensor forms
• CPP so that we that we can choose what we are linking against (@SimonClifford)
• Adapt the input.nml approach from WaveNet work with configuring for different GWPs

[jatkinson1000]

I am tackling some of this in the work I am doing towards #4
As part of this I will try and write in such a way that Forpy and TF can be easily coupled in as well for comparison.

[jatkinson1000]

As part of #4 I have brought the relevant variables for 'DaveNet' down and created an interface that allows ML 'python' routines to be called using various methods (TorchScript, Forpy).

The tensors are put into the correct shape as part of this interface and then passed out to the external routines.

The linking is done at build time using CMake with the option for -DBUILD_PT_COUPLING=ON, or -DBUILD_FORPY_COUPLING=ON rather than in cpp.

2 options have been added to the input.nml file: a path to the location of the ML model, and the file/module name of the model.

Documentation should now really be added to explain the build procedures/setup.