Seemingly fixes an issue which could cause a few pixels of the remote video in a call to be scaled out of proportion.

Updates to FMJ r20 (which reverts r18, since it causes problems -- symptoms are degraded video quality).

Works on preventing an issue with the display of remote video which could scale it down to a tiny number of pixels. Fixes multiple technical/programming errors, warnings, javadocs.

Adds log entries when a Cobri (Colibri) channel's SSRCs change. Minor corrections to comments and logging.

Removes attempts to register the speex JavaEncoder and JavaDecorer classes, since they are exluded in the default libjitsi build, and trying to register them raises exceptions.

Fixes the display of the remote video in a call which could appear as a single pixel (on Windows at least).

Improves the hotplug response times on Windows which may be noticeably prolonged due to thread scheduling in cases of Windows Multimedia reporting lack of a device driver.

Fixes inconsistent synchronization in RTPSourceStream which could lead to unexpected behavior. Enables debug output in the PortAudio binaries on Windows. Makes sure that failures to load the JNI counterparts of CoreAudioDevice will not break PortAudio and the software volume control. Fixes warnings.

Commits work in progress on utilizing Android's native hardware (and software) decoding (and encoding) via MediaCodec.

Fixs a NullPointerException in the H.264 encoder which could prevent the building of a flow graph (e.g. Android). Does not attempt to enable RTCP feedback for H.264 video on Android because that does not have a way of working there.

Does not allow the FMJ encoder and decoder classes of Opus to register with FMJ/the application unless the jnopus JNI library is loaded and appears functional. Otherwise, the Opus codec will be considered supported and advertised as such while the JNI library may not be available at all (e.g. Android).

Updates FMJ to r15: Removes a NullPointerException which is said to have been placed for the purposes of JMF compatibility but is inappropriate anyway.

Uses device UID or name (instead of index) for PortAudio CaptureDeviceInfo locator. This fixes the problem of starting/stopping the wrong device after an update of the device list (i.e. a device has been dis/connected.).

Updates audio device selection: selected device is now put on top of the device list, instead of on top of the active device list.