[ES|QL] Do not skip a remote cluster base on the query's execution time status

[fang-xing-esql]

This PR attempts to prevent a fork or subquery branch from being skipped on a remote cluster when the query's execution status is changed(to SUCCESSFUL or PARTIAL for example) at execution time, and makes sure the results are collected from all branches executed in parallel. It does not correct the execution status or metrics collected on fork or subquery.

This PR is a prerequisite for enabling CCS tests for fork and subquery (#137776). Without this change, the results of these CCS tests can be non-deterministic because some branches may be skipped unexpectedly, and cause incorrect query results. This fix can either be merged together with #137776 or submitted by itself. This change is behind snapshot and is applied to branches from fork and subquery only, so that the CCS tests for fork and subquery can return consistent and correct results. The existing CCS tests for queries without fork or subquery branches should pass.

We might need new data structure to support collecting correct execution status and metrics for fork and subquery, and it will be addressed in a separate PR.

[elasticsearchmachine]

Hi @fang-xing-esql, I've created a changelog YAML for you.

[elasticsearchmachine]

Pinging @elastic/es-analytical-engine (Team:Analytics)

[fang-xing-esql]

The two failed release tests happen on main as well, they are not introduced by this PR.

[ioanatia]

looks good! is the next step to enable the multi cluster tests running with snapshot builds for FORK and views?

[fang-xing-esql]

looks good! is the next step to enable the multi cluster tests running with snapshot builds for FORK and views?

Thank you for reviewing @ioanatia ! Yes, the next step is to enable MultiClusterSpecIT for fork, subquery and view.