lloidparts.py: update mkLLOIDmulti() to have a default h(t) gate threshold of infinity instead of None

gstlal_inspiral: update gstlal inspiral to have a default h(t) gate threshold of infinity instead of None

- also, don't forget to regenerate the combined rates after updating
  zero-lag counts

(names of people who committed ot them from git log)

- note that there's no attempt to only wipe likelihood ratios for
  inspiral<-->inspiral coincs.  you've been warned

- have encountered DAGs where somehow some trigger files didn't get
  processed.  result was a cryptic IndexError.  put in an assert to provide
  a more useful error message

accept a tuple of two values as its argument, but needing instead of list of
tuples of two values;
dict(max(self.stream_thinca.last_coincs.coinc_event_index.iteritems(), key =
lambda (coinc_event_id, coinc_event): coinc_event.likelihood)) ->
dict([max(self.stream_thinca.last_coincs.coinc_event_index.iteritems(), key =
lambda (coinc_event_id, coinc_event): coinc_event.likelihood)])

- this also replaces the approximate expression for the count at each
  iteration with the actual integral.  the weight is still placed in the
  resulting array at wherever the central value falls.

- pushes binning boundaries to [0, +inf) thereby preventing lost
  injections due to edge effects
- this also ports the process of constructing PDFs from bin counts to the
  new scheme recently introduced in pylal.  this is required in order to
  ensure the new SNR,\chi^2 PDFs are zeroed below the SNR threshold between
  the smoothing and PDF normalization steps.  (the new scheme allows
  complete customization of the counts-->PDF conversion process)

- adjustments based on recent use by me

- remove glitch model.  not what this is for.

gstlal_inspiral_pipe: fix other broken parent child relationships and add missing prior dist stats file to injections

Clustering process is naive currently, but the plan is to add a more rigorous
clustering process at this location in the future

- adjust normalization by including P(horizons).  it's a constant, so
  it's irrelevant, but now I *really* think the result is correctly
  normalized.