threads(M4): futex_wait/futex_wake, the blocking primitive

New private syscalls futex_wait(addr, expected, timeout_ns)=40 and
futex_wake(addr, count)=41. A waiter is a .blocked task tagged with
Task.futex_addr (no queue linkage); futex_wait reads the user word under the big
lock and parks only if it still equals expected, so a concurrent wake can't slip
between the check and the block. futex_wake scans the task table and readies up
to count waiters in the same address space. A timed wait also sets wake_at so the
existing wakeExpired times it out; futex_addr staying non-zero (only futex_wake
clears it) distinguishes timeout from a real wake. Waiters park in-kernel, so an
idle core still halts (no busy-wait).

runtime.Thread.Futex mirrors std.Thread.Futex (wait/timedWait/wake). thread-test
gains a futex mode: a waiter parks, the main thread wakes it (serial order
waiting/waking/woke, asserted by the case regex), and timedWait reports a
timeout.

Gate thread-futex PASS (3x); 18 guardrail cases green incl. sleep/event/ipc
blocking paths; build + host tests clean.

system/services/thread-test/thread-test.zig

@@ -117,7 +117,62 @@ fn runJoinMode() void {
     write("thread-test: join ok\n"); // the M3 verdict marker
 }
 
+// --- M4: futex mode ---------------------------------------------------------
+
+const Futex = runtime.Thread.Futex;
+
+var futex_word = std.atomic.Value(u32).init(0);
+var waiter_parked = std.atomic.Value(u32).init(0);
+
+fn futexWaiter() void {
+    write("thread-futex: waiting\n");
+    waiter_parked.store(1, .release);
+    // Block while the word is still 0; the waker sets it to 1 and wakes us.
+    while (futex_word.load(.acquire) == 0) {
+        Futex.wait(&futex_word, 0);
+    }
+    write("thread-futex: woke\n");
+}
+
+fn runFutexMode() void {
+    write("thread-futex: starting\n");
+
+    const waiter = runtime.Thread.spawn(.{}, futexWaiter, .{}) catch {
+        write("thread-futex: FAIL spawn refused\n");
+        return;
+    };
+    // Let the waiter reach its wait, then give it a beat to actually park in-kernel.
+    var spins: usize = 0;
+    while (waiter_parked.load(.acquire) == 0 and spins < 50_000_000) : (spins += 1) {
+        runtime.system.yield();
+    }
+    runtime.system.sleep(50);
+
+    // The handshake: publish the value, then wake the parked waiter.
+    futex_word.store(1, .release);
+    write("thread-futex: waking\n");
+    Futex.wake(&futex_word, 1);
+
+    waiter.join(); // returns once the waiter woke and printed "woke"
+
+    // Timeout: nobody ever wakes this word, so timedWait must report a timeout.
+    var lonely = std.atomic.Value(u32).init(0);
+    if (Futex.timedWait(&lonely, 0, 100_000_000)) |_| {
+        write("thread-futex: FAIL timedWait did not time out\n");
+        return;
+    } else |_| {}
+    write("thread-futex: timeout ok\n");
+
+    write("thread-futex: ok\n"); // the M4 verdict marker
+}
+
 pub fn main(init: runtime.process.Init) void {
     const mode = init.arguments.get(1) orelse "spawn";
-    if (std.mem.eql(u8, mode, "join")) runJoinMode() else runSpawnMode();
+    if (std.mem.eql(u8, mode, "join")) {
+        runJoinMode();
+    } else if (std.mem.eql(u8, mode, "futex")) {
+        runFutexMode();
+    } else {
+        runSpawnMode();
+    }
 }