Got a ton of benchmarking and stress testing code spit out by the
LLM. Did a not-so-thorough vetting, not ready for merge into master,
but it mostly makes sense. Needs to actually fail on the stress
tests, not anticipate failure and call it success. The benchmarks
also need a more thorough going over, in order to validate that they
make sense. Then we can work on improving our metrics.
Runtime is now configurable. Adds a tunable config for modifying the
size of pre-allocations for actor messaging channels, and for selecting
the number of threads the runtime will use.
