Support user-defined (in)equality constraints also for scipy minimisation

[thehrh]

The absence of constraints (inequality or equality) is hard-coded by PISA within _fit_scipy , see

pisa/pisa/analysis/analysis.py

Line 1726 in 41c0e89

cons = ()

and

pisa/pisa/analysis/analysis.py

Line 1826 in 41c0e89

constraints=cons,

, even though e.g. SLSQP, L-BFGS-B, or SHGO support them.

According to the documentation, inequality constraints for NLOPT can be defined as part of minimiser settings (

pisa/pisa/analysis/analysis.py

Line 848 in 41c0e89

"ineq_constraints": [

), and are converted into a callable internally (

pisa/pisa/analysis/analysis.py

Line 2446 in 41c0e89

ineq_func_params = eval(constr)

).