fix: classify numeric columns as aggregation in readAsScalar

Pie charts and scalar panels break when ClickHouse queries use custom
aggregation aliases (e.g. `count() AS total_requests`) because
readAsScalar only recognised __result_N pattern columns as aggregations,
returning columnType "group" for everything else.

Apply the same fallback logic already used in readAsTimeSeries: if a
column doesn't match an explicit marker (__result_N or a legacy alias),
infer aggregation from numeric DB type.

Closes #8844

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

pkg/querier/consume.go

@@ -292,6 +292,11 @@ func readAsScalar(rows driver.Rows, queryName string) (*qbtypes.ScalarData, erro
 		colType := qbtypes.ColumnTypeGroup
 		if aggRe.MatchString(name) {
 			colType = qbtypes.ColumnTypeAggregation
+		} else if slices.Contains(legacyReservedColumnTargetAliases, name) {
+			colType = qbtypes.ColumnTypeAggregation
+		} else if numericKind(colTypes[i].ScanType().Kind()) {
+			// Custom alias or unnamed aggregation: infer from numeric type.
+			colType = qbtypes.ColumnTypeAggregation
 		}
 		cd[i] = &qbtypes.ColumnDescriptor{
 			TelemetryFieldKey: telemetrytypes.TelemetryFieldKey{Name: name},