fix(scan-result-shaping): remove root directory entry when --strip-root is set

[abraemer]

Summary

• --strip-root now excludes the root directory entry from the files output array, matching ScanCode's skip_root=True behavior
• Fixes a duplicate-path collision bug when a child directory shares the root's name (e.g., myproject/myproject/inner.txt produced two entries with path myproject)
• Single-resource scans (one file or one empty directory) preserve the root entry, consistent with ScanCode's has_single_resource exception

Details

Previously, normalize_paths with --strip-root would keep the root directory entry in the output but rename its path to just the basename (e.g., /tmp/myproject → myproject). ScanCode instead uses skip_root=True in Codebase.walk() to exclude the root entry entirely. This divergence caused duplicate path entries in the collision case and was a general parity bug.

The fix adds a dedicated step at the start of normalize_paths that removes the root directory entry when strip_root=true and the scan has more than one resource. The path normalization loop continues unchanged after removal.

The normalize_paths_for_test signature was updated from &mut [FileInfo] to &mut Vec<FileInfo>, and a workaround in golden test utilities that re-added the root entry after normalization was removed (no longer needed).

Test coverage

• normalize_paths_strip_root_removes_root_directory_entry — root dir removed, children kept
• normalize_paths_strip_root_keeps_root_entry_for_single_file_scan — single-resource file exception
• normalize_paths_strip_root_keeps_root_entry_for_single_directory_scan — single-resource empty-dir exception
• normalize_paths_strip_root_removes_root_but_keeps_same_named_child — collision case no longer produces duplicates
• normalize_paths_without_strip_root_keeps_root_directory_entry — default behavior unchanged