Add wait event for pg_usleep() in perform_spin_delay()

The lwlock wait queue scalability issue fixed in a4adc31f69 was quite hard to
find because of the exponential backoff and because we adjust spins_per_delay
over time within a backend.

To make it easier to find similar issues in the future, add a wait event for
the pg_usleep() in perform_spin_delay(). Showing a wait event while spinning
without sleeping would increase the overhead of spinlocks, which we do not
want.

We may at some later point want to have more granular wait events, but that'd
be a substantial amount of work. This provides at least some insights into
something currently hard to observe.

Reviewed-by: Michael Paquier <michael@paquier.xyz>
Reviewed-by: Robert Haas <robertmhaas@gmail.com>
Reviewed-by: Alexander Korotkov <aekorotkov@gmail.com>
https://postgr.es/m/20221120204310.xywrhyxyytsajuuq@awork3.anarazel.de

src/backend/storage/lmgr/s_lock.c

@@ -53,6 +53,7 @@
 #include "common/pg_prng.h"
 #include "port/atomics.h"
 #include "storage/s_lock.h"
+#include "utils/wait_event.h"
 
 #define MIN_SPINS_PER_DELAY 10
 #define MAX_SPINS_PER_DELAY 1000
@@ -136,7 +137,17 @@ perform_spin_delay(SpinDelayStatus *status)
 		if (status->cur_delay == 0) /* first time to delay? */
 			status->cur_delay = MIN_DELAY_USEC;
 
+		/*
+		 * Once we start sleeping, the overhead of reporting a wait event is
+		 * justified. Actively spinning easily stands out in profilers, but
+		 * sleeping with an exponential backoff is harder to spot...
+		 *
+		 * We might want to report something more granular at some point, but
+		 * this is better than nothing.
+		 */
+		pgstat_report_wait_start(WAIT_EVENT_SPIN_DELAY);
 		pg_usleep(status->cur_delay);
+		pgstat_report_wait_end();
 
 #if defined(S_LOCK_TEST)
 		fprintf(stdout, "*");