pyfstat.py

The motivating idea here is to give the user an idea of how well the
MCMC is likely to perform: if the typical number of templates is of
order millions or more then the MCMC will work like a random template
back, jumping back and forth without finding a peak (for a long time).