- move P_instruments_given_signal() and joint_pdf_of_snrs() from far.py to
  new module

- now that the horizon distances are set to 0 in the histories when the instruments are off, populate_prob_of_instruments_given_signal() no long requires segment lists to do its job

- from method of inspiral.Data to method of lloidparts.Handler
- first step towards recording unlocked stretches as intervals of 0 horizon
  distance, ultimately allowing ranking statistic code to rely exclusively
  on horizon distance history and drop use of segment lists

- move to separate module in gstlal.stats name space
- fix "bounds out of order" with open-ended slices: when open-ended slice is beyond end of the keys currently in the tree
- also make a few adjustments:  use isinstance() to check for slices in NearestLeafTree, check for negative horizon distances in quantize_horizon_distances()
- add .weighted_mean() methods

- move some common general statistics routines from gstlal-inspiral's far.py into the gstlal.stats.__init__ module

4393

gstlal-inspiral/bin/gstlal_bank_splitter: Change naming scheme so T050017
description begins with the bin number
gstlal-inspiral/python/inspiral_pipe.py: Added function that takes a group of
T050017 files and generates a name for the collection of files that follows the
T050017 format.
gstlal-inspiral/bin/gstlal_inspiral_pipe: Took advantage of cache option on
more jobs.

Revert "lloidparts: add drop element back in. FIXME: this unlocks injection pipelines but I have no idea why"

This reverts commit d96e9fc8.

lock-up fixed by c749a20d

- probably still broken though

- it causes lock-ups.  give each orthogonal SNR stream its own segmentsrc and gate

- cache files contain URLs, not paths

- this has been fixed upstream in glue.ligolw

This reverts commit ffbe368d.

lloidparts: add drop element back in. FIXME: this unlocks injection pipelines but I have no idea why

- not a "numpy.float64" (confuses SVD bank I/O code)