Reported by Thomas Kuntz.

Because now RtpdumpStream doesn't use the rtpdump timestamp to sync
the stream. Now the sync use the timestamp of rtp packet directly, with
the timebase computed from the framerate/samplerate of the Format.

Also change the static method that create rtpdump MediaDeviceImpl :
it now takes a MediaFormat, so that it can get the clock rate of the
format (used in the timebase).

but provide a static method creating a AudioMediaDeviceImpl
or MediaDeviceImpl (for video) for convenience instead.

classes of the rtpdump package now use RtpdumpPacket, that extends RawPacket
(so the parsing is done by RawPacket).

Also fix a javadoc typo.

packet that were recorded.

Also add a custom MediaDevice for convenience.

HighestQualityRTCPTerminationStrategy now extends BasicRTCPTerminationStrategy reducing code duplication.

Renames writeRTCPFeedbackMessage() -> writeControlPayload() in various places (RTPTranslatorImpl, RTPConnectorImpl, ...)

https://github.com/jitsi/libjitsi/pull/11#issuecomment-49404492
Also fix a bug where the first frame only was send (when loopFile was true) :

FFmpeg doesn't set the number of frame contained in the IVF file, so
the IVFFileReader was reading only the first frame because it thought
it has reached the last frame. It now use the EOF to know if the end
is reached or not.

in the IVF file to sync the stream, instead of just
using the framerate of the frame.

frames for the doRead function...

Also add a custom MediaDevice for convenience.