pseudo-priors and derived parameters

[ntessore]

This PR allows priors to be declared pseudo-priors, which return a value not using some random distribution but a given function. For now, only the delta function prior is a pseudo-prior.

Additionally, parameters with a pseudo-prior are now treated as derived parameters for MultiNest (cf. #27), reducing the dimensionality of the parameter space that has to be explored.

There is a caveat: MultiNest expects the derived parameters to be at the end of the parameter list, and this is reflected in its generated files. To make sure the parameter order is correct, check out the generated .paramnames file.

[ntessore]

I'm not sure there's anything we can do to keep MultiNest's output in line with the one from Lensed, except

1. reorder the parameters in Lensed as well, or
2. write the posterior table ourselves.

Going with the first option could mean displaying two parameter tables, one for real and one for derived parameters. This would split objects apart, but make it clearer which parameters are real and which are derived.

The second choice would be nicer if we could write all output files ourselves, but right now it is not possible to access MultiNest's information on multiple modes programmatically.

@fabiobg83 @rbmetcalf What do you think?

[ntessore]

@rbmetcalf @fabiobg83 Once Travis turns green I will merge this. If the MultiNest output poses a problem in the future we can handle it then.