Add set_hyperparameters method for PopulationMCMC

[ben18785]

I think, to make things simple for running benchmarking / functional tests, we should:

1. Add another argument to set_temperature_schedule that allows setting of a schedule on the log scale. But would adding another option to an set_hyperparameter_* method complicate Change set_hyperparameters to accept a dictionary and named arguments #1320 ?
2. Allow only an integer to be passed to set the number of gradations
3. Include the initial_phase_iterations and the eta parameters of the underlying adaptive covariance method as other hyperparameters (looking at adaptive covariance, however, initial_phase_iterations doesn't seem to be given as a hyperparameter, but perhaps it should be?)

[MichaelClerx]

For now, I've only used set_temperature_schedule as is.

1. I think point 1 is part of Set temperature schedule on population MCMC to be uniform on log scale #409 . So whoever does that can have a look at this?
2. This is already possible
3. This should probably be handled whenever we allow different internal methods. Have made a ticket for that here: Allow PopulationMCMC to use different internal methods #1382

[ben18785]

Thanks @MichaelClerx -- agree with all your comments here.