test: Brush up test contracts

[slowli]

What ❔

Extracts test contracts to a separate crate with a reasonable build pipeline.

Why ❔

For now, test contracts are distributed across multiple crates (e.g., loaded using hardcoded paths in the workspace). This is not maintainable and makes the codebase harder to publish and use.

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_supervisor fmt and zk_supervisor lint.

[slowli]

The regression in the iai benchmark is entirely explained by the increased time to read system contracts (more specifically, most additional instructions are in serde_json::read::next_or_eof function, so it looks like parsing Solidity contracts in zksync_contracts). Not entirely sure which changes in zksync_contracts slowed it down; logically, the only change is removal of the load test contract.

[slowli]

This PR isn't 100% brushed up, but I think it should give a good general understanding of the approach. Essentially, building test contracts is now moved to a build script (which still relies on yarn; I haven't explored options like foundry-compilers yet), so that the building process is fully encapsulated inside the corresponding crate. IMO, this approach has significantly better in terms of DevEx than the previously used one, but there may be significant downsides I'm missing.

[github-actions]

Detected VM performance changes

Benchmark name	change in estimated runtime	change in number of opcodes executed
deploy_simple_contract	+8.7%	+0 (NaN%)
deploy_simple_contract_legacy	+7.4%	N/A

Changes in number of opcodes executed indicate that the gas price of the benchmark has changed, which causes it run out of gas at a different time. Or that it is behaving completely differently.