- there was a time when streamthinca needed to define a special class to be
  used for triggers to allow the coincidence algorithm to work, and so it
  made sense for the definition to live in streamthinca where the
  requirements were defined.  today no special class is required, the
  underlying trigger class is simply used as-is, but code like lloidhandler
  is still importing the symbol from streamthinca instead of importing it
  themselves directly.  this patch untangles this now gratuituous
  interdependency.

update O3 optimized Makefiles: bump ligo-scald to 0.3.0 and build from tarball, remove urllib3 from build

	--move template making step to svd_bank_snr.FIR_SNR.make_template()
	--calculte the template duration instead of inputting template duration
	--correct snr's epoch by adding time offset due to conditioning of templates

ChiWai Chan authored 5 years ago and Kipp Cannon committed 5 years ago

- fix peakoffset determination (get from trigger, don't try to find it in
  the SNR time series, that doesn't work if there are other peaks)

gstlal_compute_strain:  Use short linear phase filter after actuation filters to apply kappas instead of replacing entire static filter.  This preserves the quality of the high-pass filter.

binjfind needs them to identify sim_burst <-> sngl_burst coincs