Release notes for feng 2.1rc1

This release enjoys almost a full rewrite of the network layer including a more
compliant (and strict) rtsp parser, a new state machine and an eventloop driven core. The media parser layer is now using the glib threading facility to offload the io intensive task of parsing to worker threads spawned ad hoc.
The old bufferpool got replaced with a more robust implementation of shared
queues called bufferqueue, as result the shared memory access system got
replaced by a simpler system using the posix message queues. Since this feature
isn't available in certain systems (namely macosx) we are more than willing to
accept a more portable implementation using unix sockets maintaining the same
semantics.
The better structure layout made feng consume less memory per request and
overall it should work better in constrained environments than before.
Some less tested features like sctp got fixed thanks to our user feedback and
support. Now it should be considered a first class citizen along the other
transports (pending somebody willing to help us standardize it's usage).