( Dynamic Step Size ) HPD_explorer and MAP_Finder

[AdityaSavara]

For HPD_explorer, user should put in MAP.
Should have a restriction when adjusting step size something like this, because it means that the edge of the mode has been reached:
If last100 log_P has stdev of > 1E4, then take Q_step_size and divide it by 10.

[AdityaSavara]

As a related idea, to MAP_finder and HPD_explorer we should also track the logP to see when it has converged. Because if it just keeps on climbing, then we know we are not finished exploring the HPD yet (that is, we have probably not yet even approached the maximum).

[AdityaSavara]

I have found that for higher dimensional systems with sharp posteriors the only reasonable course right now is to use small dimensional slides (reduced parameter space) and modulate accept probability #17 #74 Longterm, we will need to have a dynamic step size that is parameter specific and uses modulate accept probability.

#73