fn_sig query panicks for specific malformed program.

[crlf0710]

During implementation of #84401, i found that fn_sig query panicks for a specific malformed test case:
test/ui/proc-macro/issue-59191-replace-root-with-fn.rs

In #84401 I plan to make the compilation abort after the error is raised, but it would be nice to allow it to continue.

[crlf0710]

Now #84401 is ready to land, let me think more about this.
The current process is this:

1. rustc_expand properly detected this case, and emitted the error.
2. However a DefId is still allocated for this malformed structure. But a HirId is never did.
3. rustc_resolve is able to give this DefId as resolution output, despite it's malformed.
4. Implement RFC 1260 with feature_name imported_main. #84401 Sent it to fn_sig and other queries
5. These queries tries to access the hir data with the assumption that every input local def id has corresponding hir data, leading to a index out of range.

It seems i need to break this chain somewhere in the list. Do you have any mentoring advice here, @petrochenkov?

[jyn514]

@crlf0710 I would expect anything with a DefId has a HirId - I would either abort early (before allocating the DefId) or allocate a HirId to go along with it.

[crlf0710]

@jyn514 Thanks! Let me give it a try.

[Noratrieb]

The linked PR was merged and I'm not aware of any bugs, they should be tracked separately anyways.