gstlal_ll_inspiral_event_uploader: make sure min FAR event is always uploaded above OPA; no wait time for first event below OPA

gstlal-inspiral/bin/gstlal_inspiral_create_prior_diststats : set upper limit for the mismatch range to be 30% by default

introduce pre-normalization in log-scale per SNR column for signal model and a fast path cut in LR evaluation outside of the SNR-chisq boundaries. Also, make consistent boundary for sampling and signal model construction

gstlal-inspiral/python/stats/inspiral_lr.py : introduce skip_kde attribute so that KDE is switched based on old/new signal model. This needs to be reverted and overwrite finish() for signal model when we remove the old signal model.

gstlal-inspiral/python/dags/layers/inspiral.py : make config.prior.df requirement during dag generation

off in the covariance matrix and delete df[(dc<0)]=max(df). Also
add some reference for cov matrix derivation in svd_bank.py

- the delimiter was hard-coded to ",", which all known use cases use, but it must be the same (variable) delimiter character used everywhere else.

- see https://sourceforge.net/p/ezxml/bugs/26/

Bug fix in gstlal_ll_inspiral_event_plotter and gstlal_ll_inspiral_event_uploader: change random -> numpy.random

This reverts commit 8eeaecf1