Round time to nearest second in memory storage.

When samples get flushed to disk, they lose sub-second precision anyways. By
already dropping sub-second precision, data fetched from memory vs. disk will
behave the same. Later, we should consider also storing a more compact
representation than time.Time in memory if we're not going to use its full
precision.

storage/metric/memory.go

@@ -55,7 +55,7 @@ func (s *stream) add(timestamp time.Time, value clientmodel.SampleValue) {
 	// BUG(all): https://github.com/prometheus/prometheus/pull/265/files#r4336435.
 
 	s.values = append(s.values, &SamplePair{
-		Timestamp: timestamp,
+		Timestamp: timestamp.Round(time.Second).UTC(),
 		Value:     value,
 	})
 }