Introduce query/timeout/count metric

[a2l007]

Follow-up PR to #10464.

This PR adds a new query metric query/timeout/count that represents the number of timed out queries during the emission period. Timed out query counts were previously clubbed under query/interrupted/count. Having a separate metric for timed out queries can make it a little easier for cluster operators to diagnose query issues better.

Marking this as Incompatible as query/interrupted/count no longer represents timed out query counts and user defined alerting/reports may need to be adjusted to factor in query/timeout/count along with query/interrupted/count.

It would be useful to add metricsMonitors to integration tests so that it provides better testing for PRs such as this.

---

This PR has:

• been self-reviewed.
• added documentation for new or modified features or behaviors.
• added Javadocs for most classes and all non-trivial methods. Linked related entities via Javadoc links.
• added unit tests or modified existing tests to cover new code paths, ensuring the threshold for code coverage is met.

---

[capistrant]

This looks good at initial glance. I'll test and submit a review tomorrow when I have some spare time

[nishantmonu51]

👍 , LGTM.

[capistrant]

LGTM.

reviewed as well as pulled down code and ran locally. example metric:

{"feed":"metrics","timestamp":"2020-11-12T18:53:55.392Z","service":"druid/broker","host":"broker:8082","version":"0.21.0-SNAPSHOT","metric":"query/timeout/count","value":12}