Set_caps seems to be executed *after* start, so in the previous commit audio_info will not be updated before the safety check and stumbles there. After this fix things seem to be going, and a sane value for chisq is obtained.

- move files into a gstlal/ directory so that header files can be
  #include'ed using paths that will work inside and outside the source tree

in working state

update Makefile.gstlal_feature_extractor_online to produce a separate kafka dag targetted at dedicated nodesw

gstlal_ll_feature_extractor_pipe: remove kafka/zookeeper jobs, update feature extraction options to take in psd-fft-length

- port XML trigger table reading function to ezligolw

- move functions that are only used by gstlal_burst_triggergen.c into that plugin as static functions.

- upgrade to newer autotools script

- was using LIGO_SEGMENTS_VERSION instead of GLUE_VERSION to decide when to
  print the warning about a too-old glue version

Still segfaults in the middle....

Makefile.gstlal_feature_extractor_online: rearrange where condor commands are located, add requirement to not land on specific node at LLO

Makefile.gstlal_feature_extractor_online: remove conditional for dag as it is not needed, update with influx backend for monitor jobs

gstlal_ll_feature_extractor_pipe: format output channel list in format that feature extraction jobs can read, add option to disable agg jobs

currently compiles but segfaults in the middle...

feature_extractor.py: fix issue with missing channels in feature production due to using trigger_time rather than buffer time as a timestamp

- gstlal-calibration is not touched