feat: add management canister methods for interacting with the chunk store

[hpeebles]

Description

Adds upload_chunk, stored_chunks, clear_chunk_store and install_chunked_code to the management canister API.

How Has This Been Tested?

I have tested this locally by uploading a wasm comprised of 3 chunks, the I queried stored_chunks which returned the chunks, then I called install_chunked_code which completed successfully, then I called clear_chunk_store, then I queried stored_chunks again and this time it was empty.

Checklist:

• The title of this PR complies with Conventional Commits.
• I have edited the CHANGELOG accordingly.
• I have made corresponding changes to the documentation.

[hpeebles]

This will be serialized inefficiently when using serde (it'll be fine when using candid, I think?).
We should either use ByteBuf or manually implement serialize/deserialize for this field.

[lwshang]

Sure. I noticed that you were using serde_bytes::ByteBuf. My latest commit focused on making the tests pass. And I felt that it's better to keep the types in API as native as possible, instead of forcing the users to import serde_bytes.

I will go through the API data structures to make sure that all blob: Vec<u8> are enhanced with #[serde(with = "serde_bytes")].

[hpeebles]

Yeah ideally we'd expose Vec<u8> rather than ByteBuf, but when its Vec<Vec<u8>> we can't use the #[serde(with = "serde_bytes")] trick. I think the only way to avoid exposing ByteBuf would be to manually implement serialize and deserialize.

[lwshang]

@hpeebles

For the inconsistency between the interface-spec and the replica implementation, the conclusion is that the replica will be modified to align with the spec. So we have to wait for that before we can complete this PR.

And since it will be pub chunk_hashes_list: Vec<ChunkHash> in the end, we don't have to implement the custom serde methods now.

Turning this PR back to draft now.