When no start codes are found, the offset reached by the scan
is remembered, so scanning will next start there.
However, when not specifying an aount of bytes to skip next time,
baseparse skips a default of 1 byte. This can cause baseparse to
sent the same data except the first byte, which then causes the
offset to point one past the end of the buffer.

So, set the number of bytes to skip to 0.

Note that I'm not sure why most bytes aren't skipped and the
offset set at a lowish value. It could be because a long frame
might be found with more data, but it seems like this could
be quite performance heavy, so someone who knows the format
might want to check if skipping would not be better, especially
as not skipping anything may end up causing an infinite loop.
It works here though, so it gets posted.