muxcollectpads.c: fixed typo in chain causing bpf (bytes per frame) to be set incorrectly in gst_audio_buffer_clip

- implement whitener initialization from calling code supplied PSD
- had been disabled since introduction of FIR-based whitening while we
  became comfortable with it, but no longer any reason to leave it disabled

- set to 1/4 psd_fft_length if not set by calling code unless FIR_WHITENER is turned on, in which case set it to 0

- tests on O2 spectra show that BNS horizon distances are preserved up to at least factors of 16 resolution reduction

- remove commented out and abandoned attempt at linear interpolator of PSD.

- combine adjacent segments up to +/- 1 h from the trigger to estimate the trigger rate.
- this is a work-around for the situation in which vetoes slice up the SNR stream so badly that no triggers can be collected, even though very loud injections can still be recovered (which creates apparently impossible candidates)

- the algorithm is not really in-place, but who cares, getting this to work
  gets both in-place and out-of-place working.
- the stock implementation can't be used because it relies on the logical
  invert operation which cannot be implemented, here, for segments carrying
  additional metadata.

lal_transferfunction:  numerous fixes to fir filter generation.  calibration_parts.py:  data cleaning function

- harmless in current use

gstlal-calibration:  new element lal_transferfunction to compute transfer functions between 2 or more channels in real time.

gstlal_etg: fixed bug where hdf dataset rewrites caused pipeline to crash, fixed bug where last set of hdf triggers were not being written

gstlal_etg: added hdf file saving ala h5py, added utility functions and classes to idq_aggregator.py