This fixes
* some issues with Scene strips containing audio by removing
the curpos pointer from sequence structure. (the same scene
strip can now be used in a row)
That also makes the code a lot cleaner.
* fixed a corner case on the beginning of a strip, where audio was
not mixed in, depending of current audio buffer state.
* Also: made some hardwired variables macros to enhance readability.
Problem remaining: mixing the same scene strip several times (read
put it into a stack instead of into a row) has
problems with HD-audio since the same HD-audio state structure is
used and therefore the system will seek permanently, which leads to
audio distortions...