Handle Node::Crate in generics_of to prevent ICE

[blackms]

Fixes #144888

The generics_of query did not handle Node::Crate, causing an ICE when the crate root DefId reached it during dyn-compatibility checking. The query stack was:

generics_of("") → dyn_compatibility_violations(Foo) → is_dyn_compatible(Foo) → …

The fix adds Node::Crate(_) to the existing match arm for nodes with no parent generics (alongside Node::Item and Node::ForeignItem), returning empty generics for the crate root. This is consistent with how other query implementations handle Node::Crate (e.g., resolve_bound_vars, check_well_formed).

r? compiler

[rustbot]

⚠️ Warning ⚠️

• There are issue links (such as #123) in the commit messages of the following commits.
  Please move them to the PR description, to avoid spamming the issues with references to the commit, and so this bot can automatically canonicalize them to avoid issues with subtree.
  • 95a2755

[cjgillot]

generics_of should not be called on a module or crate. Should we prefer removing the offending call ?

[blackms]

You're right — I dug deeper and found that the offending caller was already fixed by #153615, which changed note_and_explain_type_err to pass Option<DefId> instead of the raw CRATE_DEF_ID. There's also already a test at tests/ui/dyn-compatibility/ice-generics-of-crate-root-152335.rs.

So this PR is redundant. I'll close it. Thanks for the pointer!