Documentation & API for tsdate.build_parameter_grid

[hyanwong]

The docs mention a function called tsdate.build_parameter_grid, which doesn't exist. Instead, if you want to provide a bespoke prior to variational_gamma, you call tsdate.prior.parameter_grid().

We should think a little about the best API for this. In general I'm happy with the idea of providing a bespoke prior, but perhaps the names need changing somehow, and some thought put into the different between a discretised prior grid and the continuous space version which we call a parameter_grid.

Note that this may also interact with the API for specifying priors on historical samples, e.g. see #214 (comment) which proposes an API for historical samples and prior truncating. I'm not sure the API is very well thought through, so additional discussion would be useful, I feel.

[hyanwong]

Re the API, it might also be nice to have a function on the prior class that turns it into a global prior, by averaging. We could do this separately for the parameter-based prior and the nodegridvalues one.

[hyanwong]

Now documented