Support query retry for HBO

[abhinavmuk04]

Description

Implement retry query support for HBO in Presto, which retries failed queries and tries to retry if the query plan has changed. If the plan has changed it will re-execute, otherwise after the plan comparison we immediately fail it back

Motivation and Context

Many times queries fail because of cases where estimates are inaccurate and various out-of-memory, errors can occur. With this change enabled, users will be able to have an automatic retry which will determine if a retry of the query can prevent their query from failing entirely

Impact

Now when a query fails and a user reruns it, ending up in the query succeeding, with this session property enabled, we can speed up this process to ensure it automatically retries to see if HBO can help the query

Test Plan

Various created unit tests deeply testing different parts of the new and amended code's functionality. Also E2E tested feature within verified cluster

Please message me or Feilong for example of this change working!

Contributor checklist

• Please make sure your submission complies with our development, formatting, commit message, and attribution guidelines.
• PR description addresses the issue accurately and concisely. If the change is non-trivial, a GitHub Issue is referenced.
• Documented new properties (with its default value), SQL syntax, functions, or other functionality.
• If release notes are required, they follow the release notes guidelines.
• Adequate tests were added if applicable.
• CI passed.

Release Notes

* Add support for query retry when History-Based Optimization can help a failed query,
   with the ``retry-query-with-history-based-optimization`` session property :pr:`23147`

[tdcmeehan]

I understand this is draft, but what is the design or idea behind this change?

[abhinavmuk04]

I understand this is draft, but what is the design or idea behind this change?

The goal is to implement a simple version of retry query support for HBO in Presto, which retries failed queries and checks if the query plan has changed before re-executing. Currently, just building out basic portions.

[feilong-liu]

I think the plan comparison logic needs some end to end test

[agrawaldevesh]

I suggest updating the test plan in the PR description to include exactly how you verified e2e that a retry is happening for both cases where a retry should happen and where it shouldn't. Please give examples of queries.

[feilong-liu]

Can you also shadow production traffic with the feature turned on?