busctl: fix timeout calculation for "busctl monitor"

The --timeout= logic was implemented incorrectly, as it would not put a
a limit on the runtime of the operation, but only on the IO sleep.
However, spurious wakeups are possible, hence the timer would be reset
too often.

Fix that, by determining the absolute timestamp early, and checking
against that.

Follow-up for: 989e843e75
See: #34048

src/busctl/busctl.c

@@ -1268,6 +1268,9 @@ static int monitor(int argc, char **argv, int (*dump)(sd_bus_message *m, FILE *f
         if (r < 0)
                 return r;
 
+        usec_t end = arg_timeout > 0 ?
+                usec_add(now(CLOCK_MONOTONIC), arg_timeout) : USEC_INFINITY;
+
         /* upgrade connection; it's not used for anything else after this call */
         r = sd_bus_message_new_method_call(bus,
                                            &message,
@@ -1382,8 +1385,8 @@ static int monitor(int argc, char **argv, int (*dump)(sd_bus_message *m, FILE *f
                 if (r > 0)
                         continue;
 
-                r = sd_bus_wait(bus, arg_timeout > 0 ? arg_timeout : UINT64_MAX);
-                if (r == 0 && arg_timeout > 0) {
+                r = sd_bus_wait(bus, arg_timeout > 0 ? usec_sub_unsigned(end, now(CLOCK_MONOTONIC)) : UINT64_MAX);
+                if (r == 0 && arg_timeout > 0 && now(CLOCK_MONOTONIC) >= end) {
                         if (!arg_quiet && !sd_json_format_enabled(arg_json_format_flags))
                                 log_info("Timed out waiting for messages, exiting.");
                         return 0;