fix passing of kwargs to optimizers

[nden]

Closes #4611

[bsipocz]

@nden - Could you please add a changelog entry?

[nden]

@bsipocz done and merging

[ycopin]

I think these changes somehow break previous behavior, and should therefore be stated more explicitly. I used to propagate the verbosity flag disp in the kwargs (the one expected by scipy.optimize.fmin_slsqp), and this now ends up in error with:

  File "/data_sys/ycopin/Dropbox/Science/Euclid/Emilines/statistics.py", line 129, in __call__
    self.objective_function, p0, farg, **kwargs)
  File "/data/ycopin/Softs/local/lib/python2.7/site-packages/astropy/modeling/optimizers.py", line 231, in __call__
    full_output=True, **kwargs)
TypeError: fmin() got multiple values for keyword argument 'disp'

One should now use the verblevel keyword, as mentioned in the doc. Wouldn't it be more sensible to change the doc to match last version's working behavior?

[nden]

@ycopin Sorry, my bad. The goal was to use the same keyword with all fitters but I see that in fixing one problem I created a different one.
Given the state of things would it be sufficient to allow either verblevel or disp as input to SLSQP?

[ycopin]

I think it would be better to simplify the code, by allowing a single well-documented keyword, and stick to it, and letting people know, maybe with a DeprecationWarning. My 2 cents...

[pllim]

I think this should be a new issue, otherwise it is buried in a merged PR and hard to find.

[nden]

@pllim Agreed. I'll make a PR asap.