Correct issues with VideoProvider discovered via CTS tests.

While authoring the CTS tests I discovered a few issues with the
VideoProvider.
1. The VideoProvider's default handler should be on the main Looper,
similar to other Telecom components.
2. When calling IVideoCallbacks, the try/catch for RemoteExceptions was
around the for-loop which iterates over the various VideoCallbacks.  If
one was to fail, the other VideoCallbacks would not be called.
3. The mVideoCallbacks hashmap should use a ConcurrentHashMap; in CTS
tests I ran into a few concurrency exceptions related to the various
InCallServices registering callbacks while other methods on the provider
were being called.

Bug: 21802841
Change-Id: Ib0d46daf03554309044e9efaa991a15cb2c4b46b