Simplify parameter boundaries interface for fisignal

[luisfabib]

Another remnant of the old MATLAB interface. In fitsignal the lower/upper boundaries of the different parameter subsets are specified as a single keyword

fitsignal(__, lb = [lb_dd,lb_bg,lb_ex], ub = [ub_dd,ub_bg,ub_ex])

Since one seldom needs to re-define all boundaries for all subsets, it would make sense to simplify this to multiple keywords

fitsignal(__, lb_dd = lb_dd, lb_bg = lb_bg lb_ex = lb_ex)

this way a quick assignment can be done without needing to worry about the order of the parameter subsets.

[stestoll]

This sounds reasonable, and is consistent with the existing interface where the three models are already provided separately. To be completely consistent, this must be extended to the starting parameter par0 -> par0_dd, par0_bg, par0_ex. So the choice is between two extremes:

• model, par0, lb, ub
• model_dd, model_bg, model_ex, par0_dd, par0_bg, par0_ex, lb_dd, lb_bg, lb_ex, ub_dd, ub_bg, ub_ex

The second approach has the downside that it has a large number of keyword arguments. Maybe there is another way of organizing all these inputs. Group par0, lb, and ub for each model together? Or for each parameter?

[luisfabib]

After some consideration it seems that the second approach will be the most pythonian way of dealing with this. Having multiple keyword arguments with explicit names instead of a single keyword with nested lists and implicit ordering seems to align very well with the PEP20 guideline.

This will make the function call much readable only at the cost of more keywords to document. A PR will follow soon.