fix: Resolve compile errors with new nested error structure

[raymondkfcheung]

This PR fixes compile errors introduced in #7730, which adds nested error reporting and instruction index tracking to pallet-xcm, xcm-executor, and related components.

Specifically:

• Updates all usage sites to match the new nested structure.
• Ensures proper propagation of the index field in Outcome::Error and Outcome::Incomplete.
• Adjusts method signatures and error handling to reflect the new error format.

This change helps improve XCM observability and debugging by explicitly identifying the failing instruction in error contexts.

[raymondkfcheung]

Reverted it, in favour to #8549

[raymondkfcheung]

@franciscoaguirre Please let me know if you'd prefer to replace (u8, Error) with a named struct like ExecutionFailure (as shown below) in this PR, or if you'd rather have it done in a follow-up PR:

pub struct ExecutionFailure {
    pub error: Error,
    pub index: u8,
}

[franciscoaguirre]

I think a named struct is clearer. The only problem is we have so many structs named very similar and we'd have to import it everywhere. We could reuse the ExecutorError but the weight field wouldn't be useful for us here.

I would still use a named struct. It provides a much better API and the Outcome of an XCM is a somewhat public API

[raymondkfcheung]

Yes, I agree. There are quite a few similarly named structs, which can be a bit confusing, but a named struct does improve readability and long-term maintainability. I’ve merged the supporting change into your feature branch. Since the current PR still uses (u8, Error), I’ll raise a follow-up PR to introduce ExecutionFailure instead. We can always revisit the name later if we think of something more fitting.

[raymondkfcheung]

Maybe InstructionError would be a better name. It feels more focused since the struct is really about capturing an error at a specific instruction index.

[franciscoaguirre]

I think a named struct is clearer. The only problem is we have so many structs named very similar and we'd have to import it everywhere. We could reuse the ExecutorError but the weight field wouldn't be useful for us here.

I would still use a named struct. It provides a much better API and the Outcome of an XCM is a somewhat public API

[franciscoaguirre]

Thanks for doing this!