Pin db threads to cores

[coolreader18]

Description of Changes

This pins the threads that wasm instances run on, tokio worker threads, and rayon threads to specific cores, in the following arrangement:

• Module instance threads: 1/8 of available cores, with a round-robin-ish repinning system that tries to ensure that if there are >nproc/8 databases running, all the database cores will have roughly the same number of instances pinned to them.
• tokio worker threads: 4/8 of available cores, with tokio::runtime::Builder::num_threads set to the same number.
• rayon threads: 1/8 of available cores.
• tokio blocking threads: on linux, all of them have their sched_setaffinity set to a cpuset of the remaining 3/8.

API and ABI breaking changes

Expected complexity level and risk

Testing

• Bot-tested and used in production for BitCraft.

[coolreader18]

Ah. I think the failure is possibly because DynModuleHost::exit() no longer actually stops things from running.