Feature proposal: `unwrap_lamports` instruction

[joncinque]

Problem

As outlined clearly in solana-program/token#87, users want to move lamports from a wrapped SOL account into any other account. This avoids needing to create and close wrapped SOL accounts constantly.

Proposed Solution

Do the same thing as solana-program/token#87, and add a new instruction for discriminant 45 which transfers all lamports to some amount of lamports from a wrapped SOL account into any other account.

[abelmarnk]

I’ve started implementing this and have opened a draft here: #857.

[abelmarnk]

@joncinque I’m running into an issue while adding the instruction to the
processor and updating the surrounding code.

The problem comes from how the cli and rust-legacy crates depend on
solana-transaction-status:

• The CLI depends on solana-transaction-status directly:
  https://github.com/solana-program/token-2022/blob/main/clients/cli/Cargo.toml#L34
• rust-legacy depends on it indirectly through solana-cli-output:
  https://github.com/solana-program/token-2022/blob/main/clients/rust-legacy/Cargo.toml#L32
  → https://github.com/anza-xyz/agave/blob/master/cli-output/Cargo.toml#L49

solana-transaction-status itself depends on the
spl-token-2022-interface crate:
https://github.com/anza-xyz/agave/blob/master/transaction-status/Cargo.toml#L49

The workspace has a [patch.crates-io] entry that forces all uses of
spl-token-2022-interface to the local version:
https://github.com/solana-program/token-2022/blob/main/Cargo.toml#L65

Because I’ve added a new instruction variant here:
https://github.com/solana-program/token-2022/blob/main/interface/src/instruction.rs#L48

solana-transaction-status now fails to compile since it doesn’t yet handle
that variant:
https://github.com/anza-xyz/agave/blob/master/transaction-status/src/parse_token.rs#L44

If I remove the [patch] line, everything compiles and tests run, but I’m
not sure whether removing it breaks anything related to #831 (the publish
jobs fix).

What is the correct approach here?

I want to avoid undoing the fixes introduced by #831.

Sorry for the links, I included them so everything is quicker to trace.

[joncinque]

Since adding a new instruction is a breaking change, we'll need to publish a breaking change to spl-token-2022-interface, so it's fine to remove the patch for now. Thanks for checking though!

[joncinque]

Implemented in #857