Fix waiting on coroutines (#5463)

Public coroutine API returns a safe reference to specific running coroutine. Fixes bug where multiple coroutines executing same procedure would identify as same coroutine.
Greatly optimizes `alive()` (and as a result of that `wait()`) calls.
Coroutine struct is allocated together with stack as memory unmanaged by GC.

tests/coroutines/twait.nim

@@ -0,0 +1,19 @@
+discard """
+  output: "Exit 1\nExit 2"
+"""
+import coro
+
+var coro1: CoroutineRef
+
+proc testCoroutine1() =
+  for i in 0..<10:
+    suspend(0)
+  echo "Exit 1"
+
+proc testCoroutine2() =
+  coro1.wait()
+  echo "Exit 2"
+
+coro1 = coro.start(testCoroutine1)
+coro.start(testCoroutine2)
+run()