Add a capability to ModifiableGMPE to apply IMT-dependent delta to total, tau or phi

[CB-quakemodel]

We already had the ability to specify an IMT-constant scalar to any of these GMM sigma components. This PR adds the ability to specify this on an IMT-dependent basis similar to the existing capabilities for scaling the median or specifying a fixed total sigma per IMT.

Classical/case_94 is expanded to test these capabilities. Also a unit test added.

Also added a test for tabular GMM and a non-tabular GMM both with extra parameters within a mixture model (classical/case_81) for documentation purposes.

[micheles]

I don't see why you have an extra dictionary. I mean, instead of passing a dictionary {'delta': {imt: value}} you could just pass {imt: value}, then instead of the syntax add_delta_to_total_std_scalar.delta = you would just have add_delta_to_total_std_scalar =. There is no need to repeat the delta, right?

[CB-quakemodel]

I think this dictionary structure is currently necessary to how ModifiableGMPE works given it needs to read add_delta_to_total_std_vector as the adjustment/method to apply, and then delta and everything within this key is the dictionary of IMT-dependent adjustments to use within add_delta_to_total_std_vector.

The approach is consistent with other IMT-dependent adjustments e.g. https://github.com/gem/oq-engine/blob/master/openquake/qa_tests_data/classical/case_54/gsim_logic_tree_modifiable_vector.xml#L13

[mmpagani]

Grazie Chris!