vss: Use the stripped header also at stream start.

vss: Use the stripped header also at stream start.

For ogg vorbis streams the stripped header is computed when a new audio file is
opened, but chunk zero, the unmodified header, is used for the first FEC group.
This not only wastes bandwidth but might cause vss to abort because the length
of the real header was not taken into account when computing the FEC
parameters.

Solve this by introducing vss_get_chunk() which returns either the stripped
header or calls afh_get_chunk() to obtain a reference to a "real" chunk.