lib/logstorage: fix time ordering in running_stats and total_stats

[func25]

running_stats and total_stats process rows in the wrong order as _time values were compared as text instead of timestamps.

Fixes them to use the actual timestamp value when available, and to parse string _time values before falling back to text ordering (backward compat). Also keep _time in updateNeededFields so ordering still works, this is documented.

Fixes #1581

[cubic-dev-ai]

All reported issues were addressed across 4 files

<sub>Reply with feedback, questions, or to request a fix.

Re-trigger cubic</sub>

[vadimalekseev]

If I understand it correctly, this PR contains both optimizations and the bug fix, I propose to only fix the bug to simplify the changes. See this git diff that also fixes the bug, but looks simplier:

diff --git a/lib/logstorage/pipe_running_stats.go b/lib/logstorage/pipe_running_stats.go
index 76c9a2009..9e6fa931c 100644
--- a/lib/logstorage/pipe_running_stats.go
+++ b/lib/logstorage/pipe_running_stats.go
@@ -271,7 +271,12 @@ func (psp *pipeRunningStatsProcessor) flush() error {
 	for _, key := range keys {
 		rows := m[key]
 		sort.Slice(rows, func(i, j int) bool {
-			return rows[i].timestamp < rows[j].timestamp
+			v1, ok1 := TryParseTimestampRFC3339Nano(rows[i].timestamp)
+			v2, ok2 := TryParseTimestampRFC3339Nano(rows[j].timestamp)
+			if !ok1 || !ok2 {
+				return rows[i].timestamp < rows[j].timestamp
+			}
+			return v1 < v2
 		})
 
 		if needStop(psp.stopCh) {

[func25]

Yeah it is simpler, but I think the above would fix only the symptom and require the better fix anyway. Also this follows the timestamp-handling approach used by the sort pipe, so we could see it as keeping timestamp handling consistent across pipes rather than as a separate optimization.