[docdb][metrics] Add metrics for number of running flushes/compactions

[bmatican]

Now that we have the new PriorityPool for flushes and compactions, we should add specific metrics, split by type (flushes, small/large compactions):

• number of running tasks
• number of paused tasks
• number of queued tasks

[bmatican]

We currently graph these in YW: managed/src/main/resources/metrics.yml

• read/write bytes
• files per compaction
• time taken

With the additional pause/resume mechanic, there's a couple of interesting things to consider:

• the sheer counts mentioned above (running vs paused vs queued up)
• how many files are currently picked by some of these compactions (which might just be the same as the "files per compaction" metric we're already using?)
• similarly, how many bytes are currently picked?

Passing this to @rajukumaryb as he's already looking at some rocksdb work.

[bmatican]

This one is a bit less well described and we should figure out which metrics make sense. So far, I noted 3 categories above, which in turn should be split by compactions that are active vs paused (ie: pre-empted)

• number of tasks
• number of files
• data size

The priority thread pool implementation is src/yb/util/priority_thread_pool.cc

For metrics usage, you can search for examples using METRIC_DEFINE_gauge. An example of another pretty static set of gauges that we setup in the generic thread utility: src/yb/util/thread.cc

Good people references: @spolitov and @jmeehan16

[bmatican]

@arybochkin I put this on the backburner for now, to switch into some tablet splitting work