[nodes] Dynamic account resolution

[joncinque]

Problem

Codama IDLs can describe all sorts of program instructions, account types, errors, and they can describe how to generate program-derived addresses based on other accounts, but they don't specify how to derive an additional dynamic set of accounts.

For example, the SPL Transfer Hook Interface specifies how additional accounts required for a transfer are supposed to be fetched, based on account data. Other systems might have other methods of resolving additional accounts, ie. through instruction simulation.

Proposed solution

I don't have a specific idea unfortunately, but it would be good for the concept of "there may be additional dynamic accounts for this instruction" to be included in Codama IDLs.

After that, we could develop specs for account resolution schemes, which Codama could support.

[lorisleiva]

This is an old issue so I might have lost the context for this (as I'm sure we discussed this together offline before you made an issue for it haha) but could you remind me how this differs to the RemainingAccountsNode which allows arbitrary accounts to be added at the end of instructions?

[joncinque]

This came out of a discussion regarding the transfer hook account resolution, which is impossible to describe in an IDL currently. It would be nice to describe those additional required accounts using an IDL somehow.