sql: fixes statement contention count metric

[j82w]

Fixes a bug introduced in #94750 where the metric
count was counting transaction that hit contention events instead of the statement count.

closes: #96429

Release note: none

[cockroach-teamcity]

This change is 

[maryliag]

Reviewed 3 of 3 files at r1, all commit messages.
Reviewable status: complete! 1 of 0 LGTMs obtained (waiting on @DrewKimball)

[DrewKimball]

Reviewable status: complete! 0 of 0 LGTMs obtained (and 1 stale) (waiting on @j82w and @maryliag)

---

pkg/sql/executor_statement_metrics.go line 238 at r2 (raw file):

			}

			ex.planner.DistSQLPlanner().distSQLSrv.Metrics.ContendedQueriesCount.Inc(1)

Won't this count number of contention events instead of number of queries with contention?

[DrewKimball]

Reviewable status: complete! 1 of 0 LGTMs obtained (and 1 stale) (waiting on @j82w and @maryliag)

---

pkg/sql/distsql_running_test.go line 335 at r3 (raw file):

UPDATE test.test SET x = 100 WHERE x = 1;
COMMIT;
SET TRACING=off;

[nit] We're no longer setting the tracing to off afterward - is that intentional? Not sure if it matters...

[j82w]

pkg/sql/distsql_running_test.go line 335 at r3 (raw file):

Previously, DrewKimball (Drew Kimball) wrote…

[nit] We're no longer setting the tracing to off afterward - is that intentional? Not sure if it matters...

I don't think it should, but I'll add it back to be on the safe side.

[j82w]

bors r+

[craig]

Build succeeded:

• Bazel Essential CI (Cockroach)