Support time_zone on composite's date_histogram

[nik9000]

We've been parsing the time_zone parameter on date_histogram for a
while but it hasn't done anything. This wires it up.

Closes #45199
Inspired by #45200

[elasticmachine]

Pinging @elastic/es-analytics-geo (:Analytics/Aggregations)

[polyfractal]

Changes look good, but I'm a bit confused why the timezone tests in CompositeAggregatorTests pass without modification?

E.g. since DateHistogramValuesSourceBuilder would set it's own timezone before, so the shard rounding math was correct (e.g. in DateHistogramValuesSourceBuilder#innderBuild()) but the VS resolution would use null and thus the formatter would be in UTC. But with this fix the correct formatter should be used, so I'm a little surprised the tests are still passing.

[nik9000]

Changes look good, but I'm a bit confused why the timezone tests in CompositeAggregatorTests pass without modification?

I'll see what I can find out!

[nik9000]

@polyfractal it looks like we just didn't have any unit tests using time zones. I added some.

[jimczi]

lgtm, thanks @nik9000

[polyfractal]

LGTM2, thanks for adding the test :)

(The test I was looking at before was testWithDateHistogramAndTimeZone(), but I'm probably misreading what it's supposed to do, only skimmed over it.)

[nik9000]

(The test I was looking at before was testWithDateHistogramAndTimeZone(), but I'm probably misreading what it's supposed to do, only skimmed over it.)

Wow, I can't believe I missed that test! I spent a while staring at it and I think it didn't catch the problem or the change because there weren't any times that changed which interval they are part of with the time zone.

[nik9000]

Thanks @polyfractal and @jimczi !