Improve genesis state performance at startup

[fab-10]

PR description

This PR takes inspiration from #6952 to improve Besu starts after the first one, specifically when the database is present, when genesis-state-hash-cache-enabled=true, since it completely avoid to reprocess the genesis state, and in case of large genesis files, with may allocations, the startup time drops from many minutes to few seconds, and in any case the genesis allocations are not kept line in memory.

The core of the change is to make the loading of the genesis allocations lazy, so they are only parsed from file when needed, and switching to a streaming parser so they are not all loaded and kept in memory.

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

[lyfsn]

Hey, I tested the current code; here are some records:

Status	Time	Memory Allocated
first start (database empty)	14 mins	658 GB
second start (database not empty)	30 s	2.25 GB

Before your refactor, the first start of Besu needed only 7 mins and 186 GB of memory allocated.
The step in there almost completely occupies 645 GB now: https://github.com/fab-10/besu/blob/758adc7d7468e52e17942d3d221116a0ecdda008/ethereum/core/src/main/java/org/hyperledger/besu/ethereum/chain/GenesisState.java#L125

The test record is in the jfr record file; you can load it directly using IntelliJ IDEA: https://github.com/lyfsn/besu-starter/blob/main/g-1.jfr

[ahamlat]

With this PR, starting with a genesis file with more than 1 GiB took 4 minutes and 30 seconds with -Xmx16g.
Without this PR, even with -Xmx16g, the node got killed after 8 minutes.