gammapy interface: 1D spectral analysis

[HealthyPear]

This PR starts to tackle #3 by adding the 1D spectral analysis case (EBL absorption is not yet taken into account, I would leave that for a separate smaller PR).

The interface is activate by adding --use-gammapy flag
to the run subcommand and will produce an estimated SED
based on the input data which can be either:

• a set or real observations from a DataStore-compatible directory
• a set of IRFs loaded from a file

Sky models are defined from a models.yml file instead than from the assumed_model section in the global config file (a next PR will unify the two - see #24).

I plan to add also a check fo the IRF pointings with allowed maximum difference with respect to the target source,
in case the provided IRFs are not very suited for that particular observation.

P.S: still in draft because the numbers do not convince me...I also added a get_wstat_table to the interface module with which you should get the equivalent of Li&Ma bin-wise significance from the faked spectrum_dataset_on_off, but the sigma showed on the SED plot correspond to gammapy likeyhood estimation.