Histogram: Smarter bounds for bucket creation

[PSeitz]

Currently the histogram pre-creates buckets in the segment collector for the whole data range in the fast field, adjusted by hard_bounds or extended_bounds if set.

In a use case where the histogram is a sub_aggregation from another aggregation on the same field which applies the some bounds, these bounds should be forwarded.

E.g.

Range on a timestamp, one bucket per month, and inside each bucket a histogram on the timestamp.