- from singles that participate in coincs so that they can be overlayed on contour plots if desired.
- this patch does not add them to plots, that will be separate patch

- don't split out lists of (ln L, fap) pairs by instrument combo, just collect as single lists

- make bin density uniform in log up to SNRs of 1000

- disregard min value of zero-lag curve if provided.

- from 100 to 40

- the patch 753825e5 factored part of fap_after_trials() out into the function poisson_not_0() which is vectoried and so returns an array object.  this makes fap_after_trials() sometimes return float and sometimes return an array depending on which internal code path is followed.  this patch fixes.

- optionally
- refs #2555

Add several features to calibration pipeline including ability to compute factors from filters file instead of EPICS channels

- fixes return type (not dtype=object)

- on each invocation of the iterative coincidence code, a list of the singles that have not formed coincidences is returned.  this patch changes the interpretation from "singles that have not formed coincidences" to "singles that have not formed zero-lag coincidences".  i.e., singles that formed time-slide coincidences are now included so that they'll be included in the denominator histogram of (snr, \chi^2) values