Use the prior_bounds for the initialization of the MCMC walker values

This implements the use of prior_bounds on the initial MCMC walkers. This should have been implemented before because if a priors bound was given, then the initial walker states should also derive from the same bounds.

Here we initialize the 1-sigma values as (<max>-<min>)/sqrt(12), used for initializing the Gaussian distribution of the walkers in their initial state for those parameters (gains and cavity pole frequency). The other parameters are initialized with uniform distributions.