Avoid splitting large range vector aggregation.

[cyriltovena]

What this PR does / why we need it:

This PR ensure we don't split large range vector aggregation in smaller interval causing waste of resource by processing more than required.

The logic is very simple, it clamp the split by to maximum range vector found in the query.

I tested this in our dev cluster and it's way better:

before

 query="sum(rate({container=\"ingester\"}[7d]))" query_type=metric range_type=range length=1h0m0s step=1s duration=46.313273078s status=200 limit=2512 returned_lines=0 throughput=12GB total_bytes=539GB queue_time=18.756237ms

after

  query="sum(rate({container=\"ingester\"}[7d]))" query_type=metric range_type=range length=1h0m0.001s step=1s duration=25.962113827s status=200 limit=2512 returned_lines=0 throughput=4.2GB total_bytes=110GB queue_time=5.923129ms

5x less data processed, for a ~2x speed improvement.

Which issue(s) this PR fixes:
Fixes #5161

Special notes for your reviewer:

Checklist

• Documentation added
• Tests updated
• Add an entry in the CHANGELOG.md about the changes.

cc @cstyan @owen-d @slim-bean based on your previous experience.

[chaudum]

LGTM
Added suggestions for improving the function comments.

[sandeepsukhani]

Nice one! LGTM!