For debugging sampling queries, we need to expose the following:

• Number of documents sampled from the collection
• Microseconds we spent sampling the collection
• Microseconds we spent in the sampling estimator
  • I'm not sure whether this should be broken down by predicate we estimate or if we should report a total number for all time spent in sampling estimator

There may be some design work here to decide where these entries should go, for example in queryPlanner or executionStats sections.