Use ValueMap for Histogram- Throughput increased by 5x (#2033)

Author: utpilla

opentelemetry-sdk/benches/metric_histogram.rs

@@ -6,7 +6,7 @@
     RAM: 64.0 GB
     | Test                           | Average time|
     |--------------------------------|-------------|
-    | Histogram_Record               | 509.21 ns   |
+    | Histogram_Record               | 193.04 ns   |
 
 */
 

opentelemetry-sdk/src/metrics/internal/aggregate.rs

@@ -1,12 +1,8 @@
 use std::{marker, sync::Arc};
 
-use once_cell::sync::Lazy;
 use opentelemetry::KeyValue;
 
-use crate::metrics::{
-    data::{Aggregation, Gauge, Temporality},
-    AttributeSet,
-};
+use crate::metrics::data::{Aggregation, Gauge, Temporality};
 
 use super::{
     exponential_histogram::ExpoHistogram,
@@ -17,10 +13,6 @@ use super::{
 };
 
 const STREAM_CARDINALITY_LIMIT: u32 = 2000;
-pub(crate) static STREAM_OVERFLOW_ATTRIBUTE_SET: Lazy<AttributeSet> = Lazy::new(|| {
-    let key_values: [KeyValue; 1] = [KeyValue::new("otel.metric.overflow", "true")];
-    AttributeSet::from(&key_values[..])
-});
 
 /// Checks whether aggregator has hit cardinality limit for metric streams
 pub(crate) fn is_under_cardinality_limit(size: usize) -> bool {