- add optional --likelihood-file command line option, enabling the
  assignment of likelihoods to coincs at runtime

- teach Data class how to snapshot and smooth likelihood ratio data and
  pass it to StreamThinca if desired

- rename .from_filename() method to .from_filenames()

- teach StreamThinca class how to pass likelihood ratio data to
  ligolw_thinca()

- update for ligolw_thinca API change

- don't store the non-coincident sngls, just return them

- fix bug that caused some columns to be left unpopulated in rows in sqlite
  databases:  ligolw_thinca was trying to set the values after appending
  rows to tables;  fix it by letting it write into Python tables and then
  copying the rows into the database in streamthinca

- tighten up the coincidence window

- convert trials table to a stand-alone class

Excess power pipeline and associated libraries + make inclusions. Right now, the pipeline is hardcoded to a certain TF decomposition and makes it up to just before the test for excess power.

- fix inability to handle 0-length buffers

- fix typo

- don't remove coincident sngls from the background counts because
  gstlal_inspiral and friends now do this themselves

- clean up some of the StreamThinca methods
- only record sngls in the background distribution histograms if they
  didn't participate in coincs

- make inspiral.Data object invoke the streamthinca.StreamThinca object from within itself

- fix a name space bug

- add the ability to control the .so library version numbers to the
  configure and build scripts (no change to the number, though).

- in DistributionsStats.synthesize_injections(), change default verbosity to False

- port injection synthesis code from far.py
- migrate to new I/O functions in pylal.ligolw_burca

- don't "from blah import *"
- replace gstlal_element_register() with direct gojbect call

and remove gstlal.ligolw_output

- clean up appsink_new_buffer() method of Data class

- remove some things from inside the coincidence loop

- don't both shadowing the process and process_params tables:  the
  ligolw_thinca() library function does not modify the contents of these
  tables.

- remove redundant make_process_params() function

- remove unused import of gstlal.ligolw_output module

- make sure the internal table object references are references to the correct objects
- make sure to release the mutex in the event of an exception in the buffer handler
- make sure there are no duplicate offset vectors in the time slide table
- make sure the ID generator is added to the sngl_inspiral table no matter what backend we're using
- edit some comments