Add diver 1.3

[patscott]

This MR adds an interface to Diver 1.3.0.

This includes adding a new yaml option to the Diver scanner to allow the user to pass initial guesses to the scanner. The new option is now demonstrated in spartan.yaml

Also included in this MR is a new getSetParameters virtual function in the ScannerBit::BasePrior class, for checking precisely which parameters priors set, and therefore which parameters need to be specified in order to use their inverse_transform method successfully. For many priors this will just be the same as getParameters, but for the composite prior in particular things get more complicated, and it becomes necessary to differentiate between

• parameters directly set by the prior in the course of a scan, and therefore needed for any application of the inverse transform (getSetParameters),
• those whose sub-priors contribute to the definition of the composite prior (getParameters), and
• the effective parameters reported by the prior as being varied in the course of a scan (getShownParameters).

The difference between these is not important for the example in this PR, but it is crucial for the example in the matching GAMBIT_light PR.

[anderkve]

Looks good to me!

[anderkve]

Thanks for doing this so quickly, @patscott! Looks good to me, and works as expected in the tests I've done.

If I understand the complication with thecomposite prior correctly, we will need to use the getSetParameters() method also for Python scanners that need to do some inverse transforms. So I added a new propery set_parameter_names to the Python module we use for the Python scanner plugins. This propery is filled by getSetParameters in the same way as the existing property parameter_names is filled by getShownParameters(). See here:

gambit/ScannerBit/include/gambit/ScannerBit/py_module_scan.hpp

Lines 865 to 870 in d504a2f

	.def_property_readonly_static("set_parameter_names", [](py::object)
	{
	static py::list names = scanner_base::to_list<std::string>(get_prior().getSetParameters());
	return names;
	})

I'll hold off merging this until you've had a chance to give a thumbs up/down for that addition.

(The follow-up task of testing how the various current Python scanners currently work/fail with the composite prior I propose we tackle as a separate issue, to not delay this PR.)

[patscott]

Perfect, thanks!