Flux point estimator API

[cdeil]

I'm having some trouble with the gammapy.spectrum.FluxPointEstimator

For the HESS test data release I noticed that by default when one makes a FluxPointEstimator and runs compute_points it gives results where the energies are in keV (should be TeV) and column dnde_errp and dnde_errn are added but not computed, which means that when plotting the flux points, the errors from dnde_err don't show up. Also the __init__ or compute_points don't expose options what to do, e.g. whether to compute upper limits or asymmetric errors or not. As a workaround, here is what I did in the HESS data release check script:

    # We do some clean-up of the results table
    table = fpe.flux_points.table
    table['e_ref'] = table['e_ref'].quantity.to('TeV')
    table['e_min'] = table['e_min'].quantity.to('TeV')
    table['e_max'] = table['e_max'].quantity.to('TeV')
    table.remove_columns(['dnde_errp', 'dnde_errn'])
    for name in ['e_ref', 'e_min', 'e_max', 'dnde', 'dnde_err', 'dnde_ul', 'sqrt_ts']:
        table[name].format = '.5e'

I would suggest to either add more options to __init__ or compute_points, or alternatively to advertise how to use the lower-level API of doing separate computations for each point.

#1140 by @bkhelifi is related. CC @registerrier @adonath @joleroi @lmohrmann

[adonath]

Addressed by #1929, #1933 and #1934.