feat(vm): Extract VM interface to separate crate

[slowli]

What ❔

• Refactors zksync_multivm crate by extracting its interface part into a separate crate.
• Revises zksync_state / zksync_multivm uses in the workspace.

Why ❔

• This eliminates multivm dependency on zksync_state, which is quite heavyweight (depends on sqlx, rocksdb etc.).
• Makes VM interface better defined and easier to reason about.

Checklist

• PR title corresponds to the body of PR (we generate changelog entries from PRs).
• Tests for the changes have been added / updated.
• Documentation comments have been added / updated.
• Code has been formatted via zk fmt and zk lint.

[slowli]

• The changeset stats look monstrous, but most files just have imports changed.
• I plan to try moving some VM-related types from zksync_types and zksync_utils into zksync_vm_interface in a follow-up PR; this one is large as-is.

[Deniallugo]

LGTM, Glad you managed to extract it

[popzxc]

I cannot even express how happy I am to see that change, thank you! ❤️

I love the fact that traits from zksync_state were moved to the new crate. If I don't miss anything, probably it means that era-test-node no longer will have to depend on zksync_dal (which was a shame, given that it doesn't use Postgres in any way), so the benefits here expand beyond this repository :)

If we think about further steps here, I have the following thing in mind:
We probably should prepare abstractions to extract execution_sandbox from the API. I have a feeling that the abstraction for VmBootstrap I proposed in the comment may be a significant chunk of work here. If it will encompass stuff like loading system contracts, then using historical/playground system contracts in the API should work out of the box. This will be a path towards having a single interface for the VM instead of 3.5 (state keeper/vm runner, API, era_test_node) and help with (currently somewhat ugly) implementation of eth_call in the consensus module.