Fix parameter limit handling in fitting #1904
This PR fixes the parameter limit handling for iminuit and sherpa in gammapy.utils.fitting.
While working on this I noticed two other small and related issues, fixed in
For the min / max scale I wasn't sure what to put as data member and what as derived property.
I looked at http://cta.irap.omp.eu/gammalib-devel/doxygen/classGOptimizerPar.html#details and see that they have always the
I think I like that a little bit better, because it's a bit more uniform.
@cdeil I see the point of putting
As long as we only get the min / max factor and never want to set it, yes.
Yes, but it's very few and extremely simple extra lines for the two properties. I have a preference for uniformity.
But maybe let's see how the whole factor API evolves, I'll continue working on tests and changes related to factor covar and errors today.
@adonath - Could you please put parameter factor and fitting backend as one point for the call on Friday? I see that currently there's a lot missing, probably most importantly