Claude review: gpu: nova-core: emit Rust metadata for dependent modules

[Claude Code Review Bot <claude-review@example.com>]

Patch Review

**Verdict: Most complex patch in the series, has a subtle concern.**

This adds a separate `rustc --emit=metadata` invocation to produce `.rmeta` without relying on the side-effect of the main object build:

```makefile
quiet_cmd_rustc_rmeta_rs = RUSTC RMETA $@
      cmd_rustc_rmeta_rs = mkdir -p $(dir $@); \
		$(rust_common_cmd) --emit=metadata=$@ \
		--emit=obj=$(basename $@).o $<; \
		rm -f $(basename $@).o
```

Observations:

1. **Why emit both metadata AND an object, then delete the object?** The `--emit=obj=$(basename $@).o` followed by `rm -f` seems wasteful. If only metadata is needed, `--emit=metadata=$@` alone should suffice unless rustc requires an emit target for codegen to complete. If rustc does require it, a comment explaining this would help.

2. **Private directory approach**: Building into `.nova_core_rmeta/` to avoid racing with the main `nova_core.o` build is sound:

```makefile
nova_core_rmeta_private := .nova_core_rmeta/$(nova_core_rmeta)
```

However, the `if_changed_dep` tracking for the private target and then `if_changed,copy` to the stable path means two targets are tracked. This adds complexity but is the right approach for correctness.

3. **The `__modname` and `part-of-builtin`/`part-of-module` private variables** are set correctly, mirroring what kbuild would set for the real module build. Good attention to detail.

4. **The `.rmeta` build invokes `rustc` a second time** on the same source. This doubles compile time for `nova-core`. Acceptable as a temporary workaround, but worth noting the cost.

---
Generated by Claude Code Patch Reviewer