Inner instructions introspection

[rista404]

It would be great to inspect and assert what inner instructions were invoked. This is quite useful for CPI events Anchor has.

Is this possible right now? If not, what changes would it require? Happy to contribute if you can provide some guidelines.

[buffalojoec]

It shouldn't be too bad! The function is private to solana-svm, but we can just reimplement it within the library.

https://github.com/anza-xyz/agave/blob/854d9cab39fe3f4ee8970310f29c39d908b79adf/svm/src/transaction_processor.rs#L986

I'd suggest placing it behind a feature flag, calling deconstruct_transaction within the harness's process_instruction, and just adding it to the InstructionResult type.