Fix #780: rtic-sync panic

[datdenkikniet]

An explicit PR that fixes the bug but sidesteps the addition of loom to avoid the side effects it may introduce.

Fixes #780

Extracts the relevant bugfix part from #1027

A description of the path leading to the bug, which also proves that the fix actually solves the problem:

1. try_recv pushes into freeq.
2. Before popping wait_queue, an interrupt occurs.
3. send no. 1 eats up the freeq slot.
4. send no. 2 is called, pushes its link to wait_queue and goes to sleep (because freeq is now empty)
5. Control returns to try_recv, which pops wait_queue (from the newly waiting send call) without touching the free queue.
6. send is awoken at some point. Its link is popped, but the freeq is empty.