file_operations._list_files: O(n^2) duplicate-dir scan, 4-5 syscalls per entry, double sort; ignore pattern '**/.*' contradicts its own comment

File: code_puppy/tools/file_operations.py - _list_files (~lines 150-300)

Severity: Medium (performance + correctness)

1. O(n^2) duplicate-directory check. For every file returned by ripgrep, the code synthesizes parent directory entries and checks for duplicates with a linear scan over the entire results list, inside a loop over path components:

for i in range(len(path_parts)):
    partial_path = os.sep.join(path_parts[: i + 1])
    if not any(f.path == partial_path and f.type == "directory" for f in results):
        results.append(...)

On a repo with tens of thousands of files this is quadratic (each any() walks all accumulated entries). Fix: keep a seen_dirs: set[str] alongside results - O(1) membership.

2. Redundant stat syscalls per entry: os.path.exists, os.path.isfile/isdir, os.path.getsize, then os.stat again - 4-5 syscalls per file where one os.stat (with try/except) provides everything.

3. Results are sorted twice with the same _sort_key (once to build file_entries for the UI message, once to build the text output, ~lines 415 and 440). Sort once into a local.

4. should_ignore_path ignores ALL hidden files by design confusion (code_puppy/tools/common.py:437): the pattern list contains "**/.*" with the comment "# Commented out as it might be too aggressive" - but it is NOT commented out. Either remove the pattern or fix the comment; right now the comment actively misleads.

5. _list_files and _grep duplicate the rg-discovery block (shutil.which + venv probing) and the tempfile-ignore-file dance verbatim - extract a _find_ripgrep() and _write_ignore_file() helper (also reduces the two separate temp-file cleanup paths).

Filed by Zen Reviewer B (code-puppy-60635a)

Start in code_puppy/tools/file_operations.py at _list_files and _grep, then inspect should_ignore_path in code_puppy/tools/common.py. Trace the ripgrep discovery, temporary ignore-file handling, directory-entry construction, metadata lookup, and output sorting. Done means the listed performance issues are addressed, shared discovery and cleanup logic is extracted, and the '**/.*' comment and behavior agree.