dma.enable() needs to be called before setting I2S0_TCSR

See conversation https://github.com/PaulStoffregen/Audio/pull/194

Combine two loops in stereo CIC computation into one for-loop.
Add compiler optimization "unroll-loops" to isr if OS enabled. Speed
improved from 113µs to 65µs per isr call for CIC interpolation filtering
(Measured on Teensy 3.2 96MHz optimized).

PT8211 with oversampling and linear interpolation by Frank Bösing.
3 Stage CIC interpolation by Ben Rheinland