Priors for MAP estimate and HMC sampling

[cgarling]

If the templates are normalized to total stellar mass (not SFR), then setting uniform priors on their coefficients does not set uniform priors on their SFRs. These priors are not yet properly dealt with. I think the easiest way might be to allow passing template_logAge through to the methods that use priors, namely fit_templates, fit_templates_mdf, hmc_sample, and hmc_sample_mdf. You could then derive the timestep for each template and include a prior that accounts for the non-uniform timesteps.

[cgarling]

Normalizing the templates to uniform SFR yields the same results for the optimizations as uniform stellar mass normalization; closing.