Simplify thread attach/detach. Use connection specific mysys_var, rather than sharing worker thread's my_thread_var with THD.

Allow for faster creation of threads in corner cases where pool would be overloaded with long non-yielding queries. 
To allow it, change minimum of thread_pool_stall_limit to be 10 milliseconds.

Also introduce a new parameter to oversubscribe a group . Number of threads running in  parallel would be higher than it normally should, leading to thrashing, but it may improving preemptiveness, which is useful for the described corner case.