feat(rpc): implement Filecoin.StateSectorPartition

[hanabi1224]

Summary of changes

Changes introduced in this pull request:

• Implement RPC method Filecoin.StateSectorPartition
• API compare tests

     Running `target/quick/forest-tool api compare /home/me/fr/snapshots/calibnet/forest_snapshot_calibnet_2024-06-03_height_1668748.forest.car.zst --filter StateSectorPartition`
| RPC Method                          | Forest | Lotus |
|-------------------------------------|--------|-------|
| Filecoin.StateSectorPartition (145) | Valid  | Valid |

Reference issue to close (if applicable)

Closes #4380

Other information and links

Change checklist

• I have performed a self-review of my own code,
• I have made corresponding changes to the documentation,
• I have added tests that prove my fix is effective or that my feature works (if possible),
• I have made sure the CHANGELOG is up-to-date. All user-facing changes should be reflected in this document.

[LesnyRumcajs]

Could you please explain why you copied https://github.com/ChainSafe/fil-actor-states/blob/main/fil_actor_interface/src/convert.rs here? If you intend to move it out of fil-actor-states, let's have a separate PR for this.

[hanabi1224]

These methods are not publicly exported.

[LesnyRumcajs]

So right now we have two lengthy implementations doing basically the same thing. I'd rather we have only one.

[hanabi1224]

Fixed.

[LesnyRumcajs]

Thank you!

[ruseinov]

I would expect this method to be a part of fil-actor-states just like we do it for other actors. Then we don't need all of those from...* method in forest either.

[ruseinov]

Here's a PR for inspiration: ChainSafe/fil-actor-states#299

Basically we could do the same with find_sector and then just call that in Forest rpc code.

[hanabi1224]

I'd prefer getting rid of the process of updating and releasing fil-actor-states crates during the RPC development and we can backport all missing functions all at once after all RPC work is done. What do you think?

[ruseinov]

It's up to you, though this creates unnecessary confusion. I'd rather go through an ordeal of having to reference a working branch in fil-actor-states than introducing these helpers and then moving them to another repo.

This also introduces another potential problem which boils down to potential mistakes when moving the code around and ultimately requires an additional round of testing once the refactoring is done.

Which is why I'm not 100% sure this saves us any hassle, while adding to the noise in forest commit history.

[hanabi1224]

find_sector added in ChainSafe/fil-actor-states#303

I'd rather go through an ordeal of having to reference a working branch in fil-actor-states

Referencing a working branch without publishing would block cargo publish --dry-run for forest.

introducing these helpers and then moving them to another repo.

I think the ordeal is because with the current development process of fil_actor_interface, there's no way to make the crate carefully designed, when the requirement of what are needed and what are not needed is unclear, and the rest of the crates in the repo are mostly copy-pasted code. Although I really love the idea of having an interface crate, I doubt the usefulness of fil_actor_interface after spending so much maintenance effort when it's not (and not likely possible to be ) a carefully designed library.

By not a carefully designed library, I mean, e.g. when the other crates are mostly copy-pasted code, why don't we use a separate version number for fil_actor_interface instead of publishing all other crates without changing anything in most of the time

[LesnyRumcajs]

Let's add changelog entry.