Improve Conflict Detection in Parallelization by Considering Slots to Reduce False Positives

[matkt]

PR description

This update improve conflict detection during transaction parallelization by checking storage slot changes. By considering modified slots in addition to addresses, we improve the accuracy of conflict detection.

Key Changes:

• Added storage slot consideration in conflict detection.
• Reduced false positives by more accurately identifying conflicts between transactions and block state.

Fixed Issue(s)

Thanks for sending a pull request! Have you done the following?

• Checked out our contribution guidelines?
• Considered documentation and added the doc-change-required label to this PR if updates are required.
• Considered the changelog and included an update if required.
• For database changes (e.g. KeyValueSegmentIdentifier) considered compatibility and performed forwards and backwards compatibility tests

Locally, you can run these tests to catch failures early:

• unit tests: ./gradlew build
• acceptance tests: ./gradlew acceptanceTest
• integration tests: ./gradlew integrationTest
• reference tests: ./gradlew ethereum:referenceTests:referenceTests

[ahamlat]

This new conflict detection improved the number of parallelized transactions from ~40% to ~60%.
The numbers in the charts below are TPS.

Before this PR (24.10.0 running with --Xbonsai-parallel-tx-processing-enabled=true)

With this PR (running with --Xbonsai-parallel-tx-processing-enabled=true)

With more transactions executed in parallel, we noticed around 20% performance improvement compared to the last parallelization implementation. The numbers below are from instances running both EL and CL and are not using high spec flag.

We can notice also almost twice better performance when comparing to besu 24.10.0 with Teku as a CL

I don't see any overhead on the resource usage level.

[matkt]

we need to do stability tests before merging

[macfarla]

changelog entry?

[ahamlat]

LGTM, comparing with the first version of the parallelization, this PR has better performances even on 4 Cores, without more overhead. Conflict detection takes even less time than the first version.

[matkt]

changelog entry?

done