CPU Affinity support

Recent _"Best practice"_ in embedded Linux and VxWorks to achieve **deterministic** results for critical functionality is to use "core reservation" at boot and take one or more CPUs out of the general pool of CPUs used for SMP scheduling.
Threads requiring determinism then use the platform dependent API schedule themselves on the reserved core. This is a request for a thread attribute that abstracts the platform specific API for affinity as part of boost::thread (not hidden in the coroutine2 benchmarking ) https://github.com/boostorg/coroutine2/tree/develop/performance