feat(lang, cli): re-added legacy idl support under legacy-idl feature flag

[tiago18c]

Summary

Added legacy IDL code generation under legacy-idl feature flag.
Added support to the above on the CLI via the anchor legacy-idl ... command (all previous command except Build and Convert).

[vercel]

@tiago18c is attempting to deploy a commit to the Solana Foundation Team on Vercel.

A member of the Team first needs to authorize it.

[tiago18c]

@jamie-osec On the lang side should anything be done re the recent zero discriminator issues?

[jamie-osec]

Hmm, not necessarily a blocker, but I notice that idl_accounts_and_functions and __idl_dispatch (in generate_legacy_idl_mod) are basically constants, which don't use any user types or data. Maybe we can define these as regular functions in lang and just call into them?
This means they can be shown in rustdoc and are easier to work on from a dev perspective.

[tiago18c]

The #[derive(Accounts)] macro is dependent on the program ID crate::ID here which is user defined.

Is there a workaround this?
I do agree this would be better, unsure how to

[jamie-osec]

Hmm, true. Any idea about the context for

// Hacky workaround because of some internals to how account attribute
        // works. Namespaces are the root of most of the problem.

Maybe we can fix that?
Again not necessarily a blocker so we could go ahead and merge as is if there's no immediate ideas

[tiago18c]

Goes all the way back to the file creation and was never changed until it was later removed in #4252

[jamie-osec]

On the lang side should anything be done re the recent zero discriminator issues?

I think if we require buffer to be a signer we should be able to close the buffer takeover attack path. The other issue wrt hijacking IdlCreateAccount is, I think, lower impact.

[tiago18c]

Trivial and shouldn't affect anyone. CreateAccount and CreateBuffer were already done in a single transaction.

[jamie-osec]

This looks basically ready, would you mind adding some brief tests that exercise each CLI option/instruction variant, and one to test that the account takeover is closed?