Three source references still point at files the module split removed #1196

#1190 split several large files into module directories. Three places in the tree still send a reader to a path that no longer exists; following any of them lands nowhere.

Nothing catches this: a path in prose is not compiled, and check-book-api-names.sh verifies device-API names, not file paths.

Two of them need more than a new path

The sentence around the path had gone stale as well, so a sed would leave a reference that resolves and still misleads.

spine_kernel_ptx.rs cites lowering.rs:132 for propagate_kernel_attrs on the next line. That function is at line 336. A line number in prose rots on the next edit of the file it names, so naming the function in its file is the form that survives.

ref_index_projections/README.md says, in the present tense, that the Rvalue::Ref arm "has five cases", and describes case 2 walking an inner loop that breaks on non-Field projections. The arm has two labelled cases today; five is what it had before the fix that README documents. The passage is a root-cause description of the pre-fix code, so it reads as a false statement about current code.

Not included

crates/rustc-codegen-cuda/examples/atomics/README.md names intrinsics/atomic.rs and ops/atomic.rs in a pipeline diagram. Those are crate-relative shorthand, and both resolve (crates/mir-importer/src/translator/terminator/intrinsics/atomic.rs, crates/dialect-nvvm/src/ops/atomic.rs), so they are correct as written.

I also swept the rest of the tree for source paths in tracked docs and comments that do not exist. The only other hits are deliberate: synthetic fixture names inside tests (intrinsics/probes/removed.ll, intrinsics/overlay/test.toml), two paths asserted absent by cuda-intrinsics-gen render tests, and the mktemp canary in check-reserved-prefixes.sh.

I have a fix and will open a PR.