voice-duck mode

[catfact]

allow voice N to modulate its playback level as a function of it's position, and that of a second voice M.

this can be used to apply a fadeout around the clicks that would otherwise occur when M is writing and N is reading, at different rates.

[catfact]

this is working ok in PR #28. but it could use some tweaks:

• [bug] currently, the position of both reading subheads is given equal weight, even if one is stopped/silent, and duck level is applied after mixing. this means sometimes an extra duck occurs around the loop endpoint. (i think)

• [enhance] duck region could be parameterized

• [enhance] currently, duck region is in units of buffer length, not units of time (it is not scaled by rate.)

• [enhance] process each voice position in turn, then write levels, then read. this would allow ducking cycles.

• [enhance] experiment with write-duck in addition to read-duck"

  • a ducking read is non-destructive to the buffer, but it doesn't allow heads to cross each other while both are recording, without writing clicks to the buffer.
  • a ducking write would i think make this behavior clickless, but would leave smooth gaps in the buffer.