fix: Prevent false-positive deadlocks on try_lock

[ShoyuVanilla]

Fixes #376

#[test]
fn try_lock_no_block() {
    loom::model(|| {
        let lock = Rc::new(Mutex::new(0));
        let lock2 = lock.clone();

        let th = thread::spawn(move || {
            let _ = lock.try_lock(); // A
        });

        let guard = lock2.try_lock(); // B

        th.join().unwrap();

        drop(guard)
    })
}

With the above code, in some execution paths, the following happens:

1. Thread 0 tries to acquire B then switched into thread 1 before finishing it
2. Thread 1 tries to acquire A and blocks thread 0 in the following lines
   

   loom/src/rt/mutex.rs

   Lines 115 to 129 in dbf32b0

   	// Block all **other** threads attempting to acquire the mutex
   	for (id, thread) in execution.threads.iter_mut() {
   	if id == thread_id {
   	continue;
   	}
   	if let Some(operation) = thread.operation.as_ref() {
   	if operation.object() == self.state.erase() {
   	let location = operation.location();
   	trace!(state = ?self.state, thread = ?id,
   	"Mutex::post_acquire");
   	thread.set_blocked(location);
   	}
   	}
   	}

3. Thread 0 is blocked, so it cannot join the thread handle

Similar thing happens with RwLock and the simple short circuit code I added prevents such paths

[mox692]

Hi, I'm happy to review this, but it might take a bit of time since it's been a while since I last looked at the code.