New concurrency API based on destructors designed for ORC

[ringabout]

This is a draft(I will keep updating it) for new concurrency API. Everything which is not marked as [OPTIONAL] is required currently. OPTIONAL means that I won't implement it.

Sync primitives

• Unified Executors
• Atomic can only be moved and refactor
• Complete the API of Atomic
• AtomicSharedPtr and AtomicConstPtr
• SharedLock / ReadWriteLock
• template withLock(varargs[Lock], untyped) accepts arbitrary locks automatically
• Condition variables(see [std/locks]close #7998(complete condition variables) nim-lang/Nim#17711)
• Add timedWait to Condition variables
• Channel add open and close support
• Channel add select support
• Latch
• Barrier
• Semaphore
• template once(see closes #8895 add std/once nim-lang/Nim#16192), template onceThread
• Parallel Algorithms [OPTIONAL]

Complete Thread API

• Thread can only be moved(see make the copy operation of Thread an error nim-lang/Nim#17734)
• proc detachThread(Thread) operation(in contract to joinThread)
• proc getThreadId(Thread) (required by isJoinable)
• proc isJoinable(Thread)(required by JThread)
• JThread (based on destructors)
• cancel threads

Thread Pool

• Fixed size/static thread pool
• dynamic thread pool [OPTIONAL]

Context

• 'msec,'sec, 'minute, 'hour support for std/times(see custom integer literals nim-lang/Nim#17489)
• proc withDeadLine
• proc withTimeout

Event loop

• io_uring module(see async: we need io_uring for Nim #592) [OPTIONAL]

Coroutine [OPTIONAL]

stackful? -- assembly + m:n threads scheduler?
stackless? -- [cps](see nim-lang/RFCs#295)? Do we need all of them? Or based on current closure based on implementation(need cancel operations and a multi-threads scheduler)

Related materials

• Executors for C++ 20
• Executors, task parallelism, channels APIs
• A multithreading runtime for Nim
• Atomics API stabilization
• Next steps for CPS