feat: add max-queued-records-per-shard metric (cherry-pick Confluent confluentinc#905)

Cherry-pick of confluentinc#905 (author: flashmouse).

Adds a SHARDS_MAX_SIZE gauge that reports the record count in the
most-loaded shard. Useful with KEY ordering to detect hot-key
bottlenecks. Also simplifies .keySet().size() to .size().

Upstream PR: confluentinc#905

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: astubbs

parallel-consumer-core/src/main/java/io/confluent/parallelconsumer/metrics/PCMetricsDef.java

@@ -38,6 +38,7 @@ public enum PCMetricsDef {
 
     INCOMPLETE_OFFSETS_TOTAL("incomplete.offsets.total", "Total number of incomplete offsets", PCMetricsSubsystem.SHARD_MANAGER, GAUGE),
     SHARDS_SIZE("shards.size", "Number of records queued for processing across all shards", PCMetricsSubsystem.SHARD_MANAGER, GAUGE),
+    SHARDS_MAX_SIZE("shards.max.size", "The number of queued records in the shards with the most queued records", PCMetricsSubsystem.SHARD_MANAGER, GAUGE),
 
 
     //TODO: Not implemented yet - add to Metrics.adoc when implemented

parallel-consumer-core/src/main/java/io/confluent/parallelconsumer/state/ShardManager.java

@@ -83,6 +83,7 @@ public class ShardManager<K, V> {
     private Optional<ShardKey> iterationResumePoint = Optional.empty();
 
     private Gauge shardsSizeGauge;
+    private Gauge shardsMaxSizeGauge;
     private Gauge numberOfShardsGauge;
 
     private final PCMetrics pcMetrics;
@@ -302,7 +303,10 @@ private void initMetrics() {
         shardsSizeGauge = pcMetrics.gaugeFromMetricDef(PCMetricsDef.SHARDS_SIZE,
                 this, shardManager -> shardManager.processingShards.values().stream()
                         .mapToInt(processingShard -> processingShard.getEntries().size()).sum());
+        shardsMaxSizeGauge = pcMetrics.gaugeFromMetricDef(PCMetricsDef.SHARDS_MAX_SIZE,
+                this, shardManager -> shardManager.processingShards.values().stream()
+                        .mapToInt(processingShard -> processingShard.getEntries().size()).max().orElse(0));
         numberOfShardsGauge = pcMetrics.gaugeFromMetricDef(PCMetricsDef.NUMBER_OF_SHARDS,
-                this, shardManager -> shardManager.processingShards.keySet().size());
+                this, shardManager -> shardManager.processingShards.size());
     }
 }