rtpjitterbuffer: improve calculation of jitter and add more stats

Given a constant source of jitter, the total-avg-jitter will be the
"correct" amount of jitter experienced.

Comparing this number to the running average for many long-running tests,
having a slightly larger window seems to approximate the jitter much
better.

In addition we avoid weighting 0 too much when the initial calculations
are done by ramping up the window and not counting the very first
measurement.
2 jobs from jitterbuffer-big-gap in 2 minutes and 38 seconds (queued for 4 seconds)
latest
Status Job ID Name Coverage
  Manifest
passed #46990
manifest

00:00:21

 
  Build
passed #46991
build fedora x86_64

00:02:17