add keywords for setting up dynesty and nautilus

[wangbingjie]

Previously, only nlive and neff could be passed to the samplers. This update provides a quick fix that allows most of the keywords accepted by dynesty and nautilus to be passed through to the corresponding samplers.

[bd-j]

Hi @wangbingjie, thanks for this. My understanding from #348 was that the parameter inspection was supposed to automatically pass any sampling keywords that are present in run_params and known to the chosen sampling algorithm, is that wrong? That said, these would need to be added to the parser given by prospect_args.py for use in the command line.

[wangbingjie]

hum... I tried passing n_like_max to nautilus (to limit the max likelihood calls for debugging; api here) but that does not seem to do anything... my code snippet is

run_params['nested_sampler'] = 'nautilus'
run_params['n_like_max'] = 2000

output = fit_model( obs, model, sps, **run_params)

perhaps I did something dumb?

[elijahmathews]

It looks to me like the unit test failures above are due to NumPy 2.0 compatibility issues, which should be fixed now with #388 merged (#388 added NumPy 2.0 compatibility, while #390 formally removes support for NumPy 1.x). Rebasing to main should result in the unit tests passing!

[bd-j]

This looks like an issue with how the inspection is used to get sampler keywords. I think I can put a fix together that will be a bit more general.