Feature/trie values column

[Harrm]

Referenced issues

Description of the Change

• Introduce spaced batch to atomically write to several db spaces
• Put trie values in a separate column from nodes
• Switch exceptions in state sync flow to outcome::result
• Remove batchable from generic storage

Possible Drawbacks

• A migration is required

Checklist Before Opening a PR

Before you open a Pull Request (PR), please make sure you've completed the following steps and confirm by answering 'Yes' to each item:

1. Code is formatted: Have you run your code through clang-format to ensure it adheres to the project's coding standards? Yes
2. Code is documented: Have you added comments and documentation to your code according to the guidelines in the project's contributing guidelines? Yes
3. Self-review: Have you reviewed your own code to ensure it is free of typos, syntax errors, logical errors, and unresolved TODOs or FIXME without linking to an issue? Yes
4. Zombienet Tests: Have you ensured that the zombienet tests are passing? Zombienet is a network simulation and testing tool used in this project. It's important to ensure that these tests pass to maintain the stability and reliability of the project. No

[ErakhtinB]

We have to consider separate flag for each type of migration in the future

[ErakhtinB]

I think we have to check getSpace result for nullptr and print critical error if it is, same about cursor()

[Harrm]

It's the responsibility of SpacedStorage to ensure it never returns a nullptr, otherwise it would return an optional. Same for cursor.

[ErakhtinB]

How do you know that? getSpace's descripion says "return a pointer buffer storage for a space" and that's it. So it's class user's responsibility to check a method result. Same for cursor: "return kv iterator". Check how they're handled in store_impl.cpp

[Harrm]

No, we never use null smart pointers, which means if getSpace returns a shared pointer to a space, it's always a pointer to a space, as specified in its description. The way it's handled in store_impl.cpp should be fixed so it doesn't have an unnecessary check.

[ErakhtinB]

no need to check for nullptr, search_state will be checked automatically

[Harrm]

I prefer an explicit check.