[Bug] pulsar-perf outputs values in nanos with values saying "ms"

### Search before asking

- [x] I searched in the [issues](https://github.com/apache/pulsar/issues) and found nothing similar.

### Read release policy

- [x] I understand that unsupported versions don't get bug fixes. I will attempt to reproduce the issue on a supported version of Pulsar client and Pulsar broker.

### Version

2.10+, 3.x

### Minimal reproduce step

run pulsar-perf consume

### What did you expect to see?

latency values in milliseconds, as labels suggest

### What did you see instead?

```
org.apache.pulsar.testclient.PerformanceConsumer - Throughput received: 332571 msg --- 3002.897 msg/s --- 165.144 Mbit/s --- Latency: mean: 18740630.690 ms - med: 18640383 - 95pct: 20181503 - 99pct: 20482175 - 99.9pct: 20512767 - 99.99pct: 20512895 - Max: 20512895
```

value sin nanos with labels "ms"

### Anything else?

I think here https://github.com/apache/pulsar/blob/e5bd77419e91d1602731cd0c1d02a738e1b7ebc7/pulsar-testclient/src/main/java/org/apache/pulsar/testclient/PerformanceConsumer.java#L464-L472 (and possibly in other places) histogram returns latency as it tracked, in nanoseconds. The perf tool should translate these into millis, as labeled.

This is not a blocker/not a priority, logging this as "gotcha" if anyone else wonders about the mismatch between values and labels.

### Are you willing to submit a PR?

- [ ] I'm willing to submit a PR!

Start in pulsar-testclient/src/main/java/org/apache/pulsar/testclient/PerformanceConsumer.java at lines 464-472, then run `pulsar-perf consume` to compare the reported latency values with their labels. Trace how the histogram values are read and confirm that the output values are expressed in milliseconds when labeled `ms`.

Written by the indexing model from the issue text.