New candidate code for Pyffmpeg

[nouvel]

Dear Pyffmpeg developers,

I have done a complete rewrite of pyffmpeg that is compiled using
cython (instead of Pyrex).
This new version include several improvements and changes compared to
the previous version of pyffmpeg.
I plan to provide a compatibility class so that programs that were
using pyffmpeg may use the new version without
needing a complete rewrite, due to the API change.

I'd like other person interested in pyffmpeg to give me their comments
on this new / version API.

The new version includes the following changes :
- support for demuxing/decompressing multiple tracks
- audio support
- support for rescaling / arbitrary pixel formats
- use of numpy instead of PIL

I already started to use quite intensively this version on Linux
Computes, it seems so far quite stable.

I included one sample program that demonstrates a very simple audio/
video playback of a multimedia file supported by PyFFMPEG, since the
audio IO are used in blocking mode, the audio may tickle a little bit.
The examples are based on PyQT and ALSA or OSS.

I am looking forward your comments.

The new version is accessible through the CVS by the mean of the
branch called "pyffmpeg2-alpha-candidate"

Best regards,

Bertrand Nouvel