Updates to Metrics Reference

Following on from MB-55403, I'm reviewing before publishing the metrics to the Server docs, and have some issues that need addressing.

I've attached a patch based on latest neo code as a starting point: goxdcr_metrics_neo.patch

I've tried to make some of the messages more consistent in layout and terminology, and would appreciate my changes being validated for correctness, and the same being done to any others, in a similar vein.

I have questions about the stability fields - there are several marked as internal and volatile and it is not clear why - can any move to stable instead?

Regarding the labels, there are some that appear to be more like tags (at least that's what they sound like from the description), for example:

        "labels": [

            {

                "name": "docCountDiverge",

                "help": "Could cause document count between source and target bucket to be different per design",

                "added": "7.0.0"

            }

        ],

Similarly, for internalError, processLatency, etc.

Are these labels being exposed in the Prometheus metrics? What values would they have?

There are also a number of labels that I was expecting to see, for example sourceBucketName, targetClusterUUID, targetBucketName, etc. Should these not be documented?

You also are setting the version added field for labels as earlier than the version that the metric has been added in.

There are also some metrics being added with a version of pre-7.x - these need to be fixed.

Regarding xdcr_data_replicated:

The number is kept for as long as a pipeline instance is active, and resets to 0 when a pipeline is restarted. This applies to other types of \"accumulative\" counters.

Is this counter being explicitly reset when the pipeline resets?