optionally dump list of segment-ids used to process the query

[himanshug]

"/druid/v3" introduced in #3319 can handle larger context objects.

enabled by "dumpSegmentList" flag in the query context.

having the list of all segment-ids returned to user has 2 primary use-cases.

1. user can see exactly what timelines were found and not found, accordingly user can choose to use the result or error out or appropriately adjust the visualization.
2. some users have caches on top of druid that cache whole query results, they can use a hash of segments used in the cache key. users typically get the list of loaded segments from coordinator HTTP endpoints (or segment metadata query) and based on that they can check whether to use cached result or not.

i'm leaving this option undocumented intentionally for now as it should only be used with "/druid/v3" really.

[himanshug]

@cheddar has reservations against exposing this information and believes that "cache" use case should be solved by druid itself, I'm going to check enabling "result" level instead of "segment" level cache on druid brokers itself.