Update the enclave dispatcher thread to wait on a condition variable for incoming writes

[eddyashton]

The core loop for the main thread inside the enclave is reading and processing things off of the inbound ringbuffer. On SGX, where we didn't trust shared memory/standard library interaction between host and enclave, we'd implemented a spin-loop with some back-off/sleep heuristics.

This updates that to use std::condition_variable. Whenever the host writes something to the ringbuffer, it notifies. The dispatcher loop waits on this condition variable, so only wakes when (and quickly when) work is available.

As a safety hedge, and to make sure this dispatcher thread processes other work (ThreadMessaging) that could arrive outside of the inbound ringbuffer, add a timeout to the wait.