Buses

[dpwe]

Implements most of #114

This adds a new "bus" field to amy_event and struct synthinfo. amy_global now has an array of bus_state_t *bus[AMY_NUM_BUSES] each of which holds params (and pointers to private buffers) for EQ, reverb, chorus, and echo.

amy_render looks at the bus field of the "lead osc" for each note, and calls render_osc_wave with a pointer to the bus-specific buffer, thus collecting the output for each bus separately (we now have AMY_NUM_BUSES * AMY_MAX_CORES buffers).

amy_fill_buffer executes EQ, Chorus, Echo and Reverb separately for each active bus, then sums the results across buses, scaled by amy_global.volume[bus], before converting to the final output samples. So, volume=<bus0_level>,<bus1_level>,...'` is now a list.

Apart from converting all the global buffers for the FX into resources managed by a struct, the weirdest thing I had to do was to hijack the d->osc field in structure delta to hold the value of the bus (instead of the osc) for commands that are setting the params of a bus - otherwise, there's no way to combine both the parameter value with the bus to which it is meant to apply in a delta. So there's a big explicit conditional in amy_event_to_deltas_queue that treats all the EQ, echo, chorus, and reverb commands as a special-case. It's a bit clumsy, but not a bad solution. Maybe we should rename d->osc to be d->target since it's not always an osc now, but that's just to avoid misdirecting readers of the code.

This went so much more smoothly than I expected. One 4 hour session to move all the FX state into structures so they could be multiply-allocated, then another 4 hour session to actually implement the per-bus processing.