Compaction metrics are not being emitted correctly

### Affected Version

Apache Druid v0.22.1.

### Description

Hey Druids!

I'm currently using Apache Druid 0.22.1 deployed on Kubernetes (v1.18) using [druid-operator](https://github.com/druid-io/druid-operator).

Aiming to create Grafana dashboards over auto-compaction behavior, I have selected the following metrics [from the docs](https://druid.apache.org/docs/latest/operations/metrics.html#coordination):

- `compact/task/count`
- `compactTask/availableSlot/count`

After setting auto-compaction on my data sources and configuring [prometheus-emitter](https://druid.apache.org/docs/latest/development/extensions-contrib/prometheus.html) extension I was able to capture metrics about ingestion, memory and others, but those referring to compaction tasks (listed above) were always returning the value `0`.

The following print screen shows this strange behavior for `compactTask/availableSlot/count`. The same occurs for `compact/task/count` even when there are compaction tasks running.


To isolate the problem I have set `druid.emitter=logging`, but the same situation was found (logs from coordinator):


Taking a quick glance at the code I found out where those compaction metrics are included at coordinator status (at `src/main/java/org/apache/druid/server/coordinator/duty/CompactSegments.java`):


Later, those variables are emitted by a `ServiceEmitter` (at `src/main/java/org/apache/druid/server/coordinator/duty/EmitClusterStatsAndMetrics.java`):


Grepping the log right above the `globalStat` at `CompactSegments.java` I could get the value set for `numAvailableCompactionTaskSlots` and `compactionTaskCapacity` variables (logs from coordinator):


**Given this scenario, am I missing some configuration or this version is running into a race condition at coordinator's duties?**

### Appendix

Leaving some additional info here to help debugging.

#### 1. Auto-compaction config

The auto-compaction config submitted to coordinators:

```json
{
"dataSource": "TEST",
"taskPriority": 25,
"inputSegmentSizeBytes": 524288000,
"maxRowsPerSegment": null,
"skipOffsetFromLatest": "P1D",
"tuningConfig": {
"maxRowsInMemory": null,
"maxBytesInMemory": null,
"maxTotalRows": null,
"splitHintSpec": null,
"partitionsSpec": {
"type": "single_dim",
"targetRowsPerSegment": 5000000,
"maxRowsPerSegment": null,
"partitionDimension": "nice_dimension",
"assumeGrouped": false
},
"indexSpec": null,
"indexSpecForIntermediatePersists": null,
"maxPendingPersists": null,
"pushTimeout": null,
"segmentWriteOutMediumFactory": null,
"maxNumConcurrentSubTasks": 2,
"maxRetry": null,
"taskStatusCheckPeriodMs": null,
"chatHandlerTimeout": null,
"chatHandlerNumRetries": null,
"maxNumSegmentsToMerge": null,
"totalNumMergeTasks": 1,
"forceGuaranteedRollup": true,
"type": "index_parallel"
},
"granularitySpec": null,
"ioConfig": null,
"taskContext": null
}
```

#### 2. Compaction config returned by the coordinator

Using route `/druid/coordinator/v1/config/compaction`, the coordinators have returned the following `maxCompactionSlots`:

```json
{
"compactionConfigs": [
{
"dataSource": "TEST",
...
}
],
"compactionTaskSlotRatio": 0.5,
"maxCompactionTaskSlots": 10
}
```

Start by tracing compactionTaskCapacity and numAvailableCompactionTaskSlots in src/main/java/org/apache/druid/server/coordinator/duty/CompactSegments.java, then follow their emission in src/main/java/org/apache/druid/server/coordinator/duty/EmitClusterStatsAndMetrics.java. Reproduce the zero values with the provided auto-compaction configuration and logging or Prometheus emitter; done means explaining and correcting why running compaction tasks are reported as zero.

Written by the indexing model from the issue text.