Add time-series query rewrite to push $limit into the PlanStage so that we don't fetch buckets that are not necessary

It is safe to assume that every bucket has at least one measurement. So we can rewrite an [{$_unpack}, {$limit: x}] pipeline as [{$limit: x}, {$_unpack}, {$limit: x}] . Similarly we can push the limit values inside $sort when we do $sort pushdown