[CM] LADSPA implementation

[e deleflie]

Sorry to be spamming the list with my problems, but I think I've
worked out a few things about how the LADSPA implementation works. I
hope my analysis is correct (it may not be).

- LADSPA output gets written to the same buffer as the input  (when a
different sound is selected, the output still goes back into the
un-selected input buffer).
- When using a reader (with make-sample-reader) the output gets
_added_ to the input buffer (so both the result and source are there).
This does not seem to happen when not using the readers (but then
temporary files dont get cleaned up in the /tmp directory). Can I
switch this off?

Feels like perhaps there should be an other parameter in the
apply-ladspa function, being the name of the output sound .....either
that or apply-plugin writes to the currently selected sound.

That said, I do seem to remember reading somewhere that LADSPA plugins
should write their result to the input buffer ... I could have that
wrong. How does an object work that has 1 input and 2 outputs? (like a
mono to stereo effect).

Etienne