ESQL: Timezone support in DATE_TRUNC, BUCKET and TBUCKET

[ivancea]

Timezone settings (Both the request parameter and the SET setting) are snapshot-only.

Add timezone support to DATE_TRUNC, BUCKET and TBUCKET.

The current timezone is already in the Configuration. This PR:

• Uses the configuration ZoneId from those functions
• Adds unit tests for edge cases
  • As BUCKET and TBUCKET depend on DATE_TRUNC, DATE_TRUNC tests are reused
• Fixed Rounding for date intervals with variable timezones not taking the amount of units (e.g. 5 days) into account
• Adds CSV tests for common cases

For reviewers

Please take a good look at the expected results of DATE_TRUNC in DateTruncTests, and comment any interesting or edge case you can think of!

[bpintea]

LGTM.
Could we add some REST tests with timezones set as JSON field / query parameter and maybe compare with another run of same query SETting it?

[bpintea]

I guess you could also have:

Suggested change

	DATE_TRUNC_TIMEZONE_SUPPORT(Build.current().isSnapshot()),
	DATE_TRUNC_TIMEZONE_SUPPORT(SET_COMMAND.isEnabled()),

if you want to link them. But then you could also introduce two caps, if you want to isolate the SET-related logic. Just some thoughts, it can stay as is.

[ivancea]

@bpintea Tests for request/SET settings were added here, if that's what you mean: https://github.com/elastic/elasticsearch/pull/136748/files#diff-8e410478016b271c9f44f50aa2d9e5f911135694d0ac6a733e2bce7d219d23e4

Do you think that's enough, or are you thinking of some other case? Thanks!

[bpintea]

My idea was to add some x-pack/plugin/esql/qa/server/single-node/src/javaRestTest/java/org/elasticsearch/xpack/esql/qa/single_node/RestEsqlIT.java tests that would run the same query twice:

• once only setting the timezone JSON parameter
• once only within the query, SET time_zone=...,

and then comparing the results. For some randomized TZs.
But maybe we have tests already that insure that the Configuration ends up with the same TZ value already.

[luigidellaquila]

LGTM

[ivancea]

@bpintea Nice idea; just added a test comparing the setting methods (+ the bucket/date_trunc/tbucket functions for extra checks)