Add support for OTel Java runtime metrics

[liustanley]

What does this PR do?

• Moves runtime metric mappings into separate file
• Adds mappings for Java runtime metrics
• Removes metricType field from runtimeMetricMapping, changes code to use incoming metric type instead
• Refactors runtime metric attributes to support metrics that map from multiple attributes
• Changes mapping code to remove attributes that were used in the mapping process (i.e. when jvm.heap_memory is mapped, it doesn't copy the "type=heap" attribute from the OTel metric)
• Support mapping histogram runtime metrics

[dineshg13]

Can we also add GC metrics ?

https://github.com/open-telemetry/opentelemetry-java-instrumentation/pull/6964/files#diff-bb4826b5d07256ae1befe69927b73586a37da8bb6dcdfa30a3ba18e5c828b55b

[dineshg13]

Looks good. Added a small comment

[sirianni]

The aliasing of process.runtime.jvm.gc.duration to jvm.gc.parnew.time breaks setups that have both legacy Java apps using JMX and new Java applications using OTel due to an incompatibility in metric type (distribution vs. gauge).

In these scenarios, the jvm.gc.parnew.time metric metadata on the Datadog backend gets changed from a gauge to a distribution which causes the JMX-fetched gauge variant of jvm.gc.parnew.time to be completely unqueryable (disappears from all dashboards and monitors).

CC @mx-psi

[mx-psi]

@sirianni can you file a support ticket for this? Please share the case number here and I can loop in the relevant people :)

[sirianni]

Please share the case number here and I can loop in the relevant people :)

https://help.datadoghq.com/hc/en-us/requests/1434627