- also fix min_ratio to 0.1
- and move some arithmetic out of the loop

- how handled by middleware in pylal.  faster

- also allow program to be restarted from a saved chain

- switch to storing logs of SNR PDFs, too

- old matplotlibs won't plot these

- need to include a representation of the horizon distance so that multiple
  PDFs don't end up with the same names

- use itertools.islice() to draw samples

- also comment the code.

far.py: only do really careful CDF checking for the combined distributions, the individual detector distributions are just for diagnostics and can sometimes come out funny. It shouldn't cause the program to crash though

gstlal_inspiral_create_prior_diststats: broaden the allowed chisq values for signals from a mismatch of 0.1 to 0.2

gstlal_inspiral_pipe: Add another create_prior_diststats to make SNR pdfs, support far away injection files being added at the last minute and fix a few minor aesthetic issues

- try removing frequency components below the timescale of the whitener's
  FFT block size from the input test stream in test 01a since the whitener
  is going to be a high-pass filter at that frequency.  I've seen the test
  fail from time to time due to what appears to be very low-frequency
  drifts in either the input or output, and I hope this will make those
  failures go away.

- don't understand, but pygobject has some sort of complaint about the type
  of the bit masks on older systems that I don't see on my own.  this fixes
  it (at least on the current Debian reference OS).

- we don't try to do anything fancy, we just ignore the version column
  altogether.  we also ignore the commen column for good luck

- deprecated function no longer exists, use code in glue.ligolw instead