[Metrics] Make Histogram Buckets Configurable

**Is your feature request related to a problem? Please describe.**
The current Pulsar metrics implementation has very fine grained histogram buckets. This leads to more metrics data, which means more compute to calculate the metrics, more memory to store them, and then more cost downstream on the service storing the historical metrics. Since different installations have different requirements, it seems to me that we could make these buckets configurable.

**Describe the solution you'd like**
Make the histogram buckets for metrics configurable. A user would simply provide a comma delimited list of bucket values. We would need several new configuration values for each type of histogram that is to be configured. It looks like the affect values are: `ENTRY_LATENCY_BUCKETS_USEC`, `ENTRY_SIZE_BUCKETS_BYTES`, and `WRITE_LATENCY_BUCKETS_USEC`. If we're concerned about proliferation of configurations, we could combine the two latency related bucket arrays in a single shared configuration.

**Describe alternatives you've considered**
Other solutions might include server side filters by metrics providers. However, that would still mean that Pulsar has to calculate the metrics, which would be unnecessary.

**Additional context**
Here are two examples of the current bucket definitions:

```java
public static final long[] ENTRY_LATENCY_BUCKETS_USEC = { 500, 1_000, 5_000, 10_000, 20_000, 50_000, 100_000,
200_000, 1000_000 };
public static final long[] ENTRY_SIZE_BUCKETS_BYTES = { 128, 512, 1024, 2048, 4096, 16_384, 102_400, 1_232_896 };
```
They have 9 and 8 buckets, respectively. The default values for these configurations would be the current values to ensure backwards compatibility.

If this feature is accepted, I will follow up with a proposal to add a similar configuration option to Apache Bookkeeper.

I'm happy to help contribute this feature.

Start by locating the Java metrics implementation and the ENTRY_LATENCY_BUCKETS_USEC, ENTRY_SIZE_BUCKETS_BYTES, and WRITE_LATENCY_BUCKETS_USEC definitions. Trace how these buckets are exposed through configuration and used to calculate metrics. Done means comma-delimited bucket settings work for the requested histograms and default to the current values for backwards compatibility.

Written by the indexing model from the issue text.