rtpjitterbuffer: rtp_jitter_buffer_calculate_pts should not reset in all cases (especially on (old) rtx packets)

While debugging some issues I noticed some strange behavior that happened occasionally. After attaching with gdb and continuing executing after a few seconds, I noticed that sometimes playback would resume and then shortly after completely be out-of-sync (permanently). I traced the problem to old rtx packets being processed, which erroneously caused it to trigger a resync to this old rtx packet due to "backward timestamps at server, schedule resync".

I fixed the issue by:

1. Add a new argument gboolean is_rtx to rtp_jitter_buffer_calculate_pts.
2. In rtp_jitter_buffer_calculate_pts, if is_rtx is TRUE for the "rtp delta too big, reset skew" and "backward timestamps at server, schedule resync" cases, then return GST_CLOCK_TIME_NONE and do nothing. I believe the other resync cases need to be triggered regardless.
3. Add a check for !GST_CLOCK_TIME_IS_VALID (pts) to where rtp_jitter_buffer_calculate_pts is called, and in those cases do a goto finished; to drop the packet.

This change appears to consistently fix playback after briefly pausing the process with gdb.