libc's __std_atomic_wait_direct_32 and __std_atomic_wait_direct_64
import these functions, so a libc.prx that contains the C++20 atomic
wait wrappers does not load without them. Wait32 compares a 4-byte
value and Wait64 the full 8-byte value. Both share the waiter queue
of sceKernelSyncOnAddressWait, so sceKernelSyncOnAddressWake releases
waiters of every size on the address. A null address, or one not
aligned to the value size, throws. The by-value comparand, the return
values and the 8-byte alignment of Wait64 are assumed and recorded in
TechnicalDebt.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>