Switch to a faster thread-safe FIFO - std::sync::mpsc is showing up in profiling as a bottleneck.

[mitchmindtree]

Currently the mpsc::SyncSender used to send audio monitoring data back to the GUI thread is showing up as a clear bottleneck. The profiling suggests that there's some inner lock contention going on that is causing delays that even exceed the cost of the FFT calculations. It might be worth looking into the crossbeam lock-free queues to solve this.