Refactor class hierarchy of NB2WProduct

[ferrigno]

This solves issue #129

[dsavchenko]

When I wrote this class, it was not intended to be directly initialised, serving as something like an abstract base class.
But if it actually works for the general-purpose NumpyDataProduct, then OK, let's use it like this.

There is some violation of the logic, however. The oda:DataProduct ontology class is a parent for every kind of products, not only NumpyDataProduct-based. So maybe it makes sense to introduce some term like oda:NumpyDataProduct to the ontology and have it as a type_key here. But in this case, python class hierarchy is different from ontology class hierarchy. Reconciling this will require some more involved refactoring.

To summarise, having this as a hotfix for the specific issue is OK for me, but I would let @volodymyrss and @burnout87 comment on this

[dsavchenko]

As a sidenote: I don't like the name NumpyDataProduct neither in the code nor in the ontology. I would have named it somehow like FITSDataProduct

[ferrigno]

I will try to discuss with @burnout87 on Monday. An alternative would be to introduce a derived class for the Fits(Numpy)DataProduct to preserve the logic or to make a check in nb2worflow for the attribute existence

[burnout87]

Having a dedicated class for a Numpy-based product makes sense, this would allow us to have the NB2WProduct class to be a proper base class, numpy-agnostic, with the type_key defined in this PR. Then, a derived class specific for Numpy-based data-products could be defined without breaking any logic, assigning its own dedicated ontology term. This can be done in a dedicated PR.

To add: I would avoid renaming the classes, especially those in the oda_api as this could potentially break the logic in many places.

[ferrigno]

@burnout87 why don't you use this PR to implement such a class without my hack ?

[ferrigno]

You can use JEM-X expert test_notebook for spectra to test this (I have set appropriate parameters)

[ferrigno]

As soon as Denys approves it, could you merge? It would be nice to have notebooks finally working also for the outputs.
I think that the guide should be updated in the output definition, as to have full functionality, one needs to use proper DataTypes and DataType annotations.