gopls/internal/telemetry: log timestamp when test fails

We observed a couple of test failures. My current hypothesis is
that the bug is in how countertest.ReadCounter and
countertest.ReadStackCounter read the data. They try to access
the counter values by reading the contents of the mapped file on
disk and parsing them with counter.Parse. They call 'file.rotate'
before reading the files to ensure the file is mapped. However,
if file.rotate ends up actually rotating the underlying file
(due to crossing the UTC day boundary), the counter data may be
flushed out and reset. As we are moving to 'week' instead of 'day'
for file rotation, I don't know how frequently we will encounter
this issue though if this was the root cause. Log the timestamp
when the test fails, to evaluate the hypothesis.

For golang/go#62161

Change-Id: If00b0788b6927640a979d17f8fe888232d27ea09
Reviewed-on: https://go-review.googlesource.com/c/tools/+/523795
Run-TryBot: Hyang-Ah Hana Kim <[email protected]>
TryBot-Result: Gopher Robot <[email protected]>
gopls-CI: kokoro <[email protected]>
Reviewed-by: Jamal Carvalho <[email protected]>

Author: hyangah

gopls/internal/telemetry/telemetry_test.go

@@ -12,6 +12,7 @@ import (
 	"strconv"
 	"strings"
 	"testing"
+	"time"
 
 	"golang.org/x/telemetry/counter"
 	"golang.org/x/telemetry/counter/countertest" // requires go1.21+
@@ -58,6 +59,7 @@ func TestTelemetry(t *testing.T) {
 		count, err := countertest.ReadCounter(c)
 		if err != nil || count != 1 {
 			t.Errorf("ReadCounter(%q) = (%v, %v), want (1, nil)", c.Name(), count, err)
+			t.Logf("Current timestamp = %v", time.Now().UTC())
 		}
 	}
 
@@ -69,6 +71,7 @@ func TestTelemetry(t *testing.T) {
 	}
 	if len(counts) != 1 || !hasEntry(counts, t.Name(), 1) {
 		t.Errorf("read stackcounter(%q) = (%#v, %v), want one entry", "gopls/bug", counts, err)
+		t.Logf("Current timestamp = %v", time.Now().UTC())
 	}
 }