gstlal_bank.xml.gz: I am deleting this. I don't know where it came from but it has the the wrong number of templates (it isn't the hyperbank) ligolw_print -t sngl_inspiral gstlal_bank.xml.gz | wc = 654261 which is not 677,000

"*1_snr", matching what was done pre-itacac

inspiral.py: Create sub-threshold trigger to add to gracedb uploads if
we have snr from an instrument that didn't produce a trigger above peak
coincident with the candidate in question. Also, only upload one snr
time series per trigger (including the sub-threshold trigger), not an
snr time series for each ifo for each trigger. After this patch, gracedb
uploads should look almost identical to what they were before switching
to itacac

- this patch fixes ID reconstruction but doesn't get the column names quite
  right.  they're close to being correct, and there is a bug in glue that
  makes it inconvenient to get them any closer, so we'll give this a try
  and see if downstream consumer codes care.  note that ligolw_no_ilwdchar
  does *not* care:  with this patch, file --> ilwdify -->
  ligolw_no_ilwdchar --> file is now an identity transformation.  also note
  that glue's normal document parsing logic does not care:  ligolw_print
  is happy with the result of this transformation.  it's not impossible,
  however, that somebody has written some custom code that adds additional
  safety checks and will notice the column names.

- don't use InterningRowBuilder

currently compiles but segfaults in the middle...

calibration_parts.py:  remove_lines_with_witnesses():  remove initial samples to avoid GSL warnings.

gstlal_compute_strain:  Allow calibration lines to be subtracted the same way that power lines are if we want.

- update to record its current state

- we still really need something even newer but this at least updates what
  the package claims it needs to the newest currently available on the LDG
  clusters

- bump min glue version to 2.0.0

lloidhandler.py: increase rate at which state data/ram history gets pushed into Kafka to 1 Hz, removing checkboard effect in heatmaps

gstlal_compute_strain:  Added option to choose the best strain channel by comparing CALIB_STRAIN to CALIB_STRAIN_CLEAN and only write out one channel.

 - tweak ligo mass model by making the bns broader, the power law for bbh steeper, and redcucing the relative BNS rate
 - also changes the assumption about template ids being from 0 to len(template bank)-1

- reduce the range of numbers used for IDs because they're being stored in
  4-byte floats

- don't construct an intermediate list object when not needed

- assign IDs randomly so that HDF5 mass models cannot be used with the
  wrong template banks:  make the odds of template IDs matching if the
  wrong files are mixed small.