[andreaTP]

Since #237, CI jobs on macOS with JDK 17 or 21 hang in MemoryTest until GitHub's 6 hour timeout, on PRs and on main.

waitAndNotifyWithAtomicsOnSameWord ran a task that never ends on the ForkJoin common pool, submitted its waiter to the same pool, then spun with no deadline until the waiter was parked. On JDK 17 and 21 the common pool, which has two workers on the 3-core macOS runners, can leave that waiter queued while its idle worker stays parked. A pool dump at the hang shows submissions = 1, one worker spinning and the other WAITING, so the test spins forever. Linux and Windows runners have 4 cores, which is why only macOS hangs.

The fix runs both tasks on two threads of their own. The memory code itself is unaffected.

Running MemoryTest alone pinned to 3 cores (taskset -c 0-2):

	JDK 17	JDK 21
before	hung in 3 of 6 runs	passed 5 of 5 here, but the same loop in jshell hung after 3,719 rounds
after	passed 10 of 10	passed 10 of 10

The CI job also gets timeout-minutes: 60, about twice its usual duration, so a hang fails within the hour instead of holding a runner for 6 hours.

[andreaTP]

Fix #247