Cache codes using KVCachedFile

[ihtefyw]

In the MPT implementation, all codes are kept in memory in a hash map. This increases memory consumption significantly and can possibly crash Carmen.
This PR adds a caching layer on top of the codes by using the KVCachedFile. Most of the existing code logic was generalized and moved to OffsetFile, a KVFile which stores the offset of each element in memory.
The cache is set to 20,000 codes + 2,000 for the flush buffer, which equals to a worst case memory usage of ~1GB.

Notable changes:

• Because now getting a code could result in a file lookup, GetCodeForHash and GetCodes return an error now, and the interface and implementations have been updated accordingly.
• getReferencedCodes in database/mpt/io/live.go was still materializing the whole code file in memory. It has been modified to write the codes while scanning them, and renamed to writeReferencedCodes

Closes https://github.com/0xsoniclabs/sonic-admin/issues/844

[Copilot]

Pull request overview

This PR introduces an on-disk caching layer for contract bytecode in the MPT implementation by replacing the in-memory code map with common.KVCachedFile, reducing steady-state memory pressure at the cost of possible file-backed lookups. It also updates the GetCodeForHash API to return an error to reflect potential I/O failures.

Changes:

• Replace in-memory code storage with KVCachedFile-backed storage with configured cache and flush-buffer thresholds.
• Update GetCodeForHash to return ([]byte, error) and propagate errors through MptState, ArchiveTrie, and export/visitor flows.
• Update and extend tests/mocks to validate error forwarding and persistence behavior.

Reviewed changes

Copilot reviewed 8 out of 10 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
go/database/mpt/verification_test.go	Adjusts verification setup for the new codes file behavior.
go/database/mpt/state.go	Updates LiveState/MptState code retrieval API to return errors.
go/database/mpt/state_test.go	Updates tests and adds cases ensuring codes read errors are forwarded.
go/database/mpt/state_mocks.go	Regenerates mocks for the updated GetCodeForHash signature.
go/database/mpt/io/live.go	Updates visitor/export path to handle GetCodeForHash errors.
go/database/mpt/io/live_mocks.go	Regenerates mocks for updated visitor interface.
go/database/mpt/codes.go	Replaces in-memory codes map with KVCachedFile and updates read/write helpers.
go/database/mpt/codes_test.go	Updates/extends tests for cached-file-backed codes persistence and lookups.
go/database/mpt/archive_trie.go	Updates archive trie code retrieval to propagate errors.
go/database/mpt/archive_trie_test.go	Adds tests validating error forwarding from LiveState/GetCodeForHash.

Files not reviewed (2)

• go/database/mpt/io/live_mocks.go: Generated file
• go/database/mpt/state_mocks.go: Generated file

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 8 out of 10 changed files in this pull request and generated 7 comments.

Files not reviewed (2)

• go/database/mpt/io/live_mocks.go: Generated file
• go/database/mpt/state_mocks.go: Generated file

[Copilot]

Pull request overview

Copilot reviewed 8 out of 10 changed files in this pull request and generated 3 comments.

Files not reviewed (2)

• go/database/mpt/io/live_mocks.go: Generated file
• go/database/mpt/state_mocks.go: Generated file

[HerbertJordan]

Some initial feedback covering things that I think should be addressed before merging this. The changes may also be substantial enough that looking through the current tests in detail is premature, so I skipped that for now. Let me know what you think.

[HerbertJordan]

Integration looks fine, but some functions still materialize the full code map. Those should be eliminated.

I did not check the unit tests.