Reduce lock contention on transaction pool when building a block

[fab-10]

PR description

Currently when Besu builds a block it takes and keep the lock on txpool, until the selection of txs to include is over (and this could be longer of the block creation timeout, since the tx that is processing across the timeout need to complete before releasing the lock) and that could delay the start of the next block building in case of small networks where single nodes could build many blocks in a row, for example L2 sequencers, but the optimization also benefit normal use cases, since incoming txs can be processed and added to the txpool concurrently to a block creation task.

Below there are 2 profiling taken, from a sequencer continuously building blocks, before and after applying this PR, selected is the layered txpool, and you can see how the Total Blocked Time went down from >10min to millis.

Before

After

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