Clarify hmatrix-gsl's SimulatedAnnealing API

[eltix]

Issue description

Thank you for the great work, we've been using simanSolve with great success so far!

We did notice some details in the API that could be improved though:

1. The n_tries attribute of SimulatedAnnealingParams is completely unused. It is used by gsl's function gsl_siman_solve_many but hamtrix-gsl's simanSolve function interfaces with gsl_siman_solve. We found it particularly confusing when we noticed that changing the value of that parameter did not have any impact onto the number of energy evaluations (and therefore no impact at all on computational time). The API would be much clearer if the parameter was removed from SimulatedAnnealingParams.

2. Similarly the distance function or "metric definition", the 6th argument of the simanSolve function, is completely unused. Again the API would gain in clarity by removing the argument altogether or at least clarify its uselessness in the documentation of the function.

Thanks again for implementing the GSL bindings!