fix: filter venv/cache/build noise from changed files, reindex all commits

Added _is_noise() filter to core/diff_reader.py to strip venv/, node_modules/,
__pycache__/, .pyc and other build artifacts from the changed files list before
they reach the LLM or appear in the build log. Imported the same filter in
scripts/reindex.py and reindexed all commits — ffaff32 (the venv leak commit)
now shows 3 clean files instead of 688+ noise entries.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Author: chanakyavasantha

core/diff_reader.py

@@ -1,5 +1,37 @@
 import subprocess
 
+# Directory prefixes that are never meaningful to track as feature files.
+# These are dependency/build/cache directories that can appear in diffs
+# when .gitignore wasn't set up yet or was misconfigured.
+_NOISE_PREFIXES = (
+    "venv/",
+    ".venv/",
+    "env/",
+    ".env/",
+    ".gitmind/venv/",
+    "node_modules/",
+    "__pycache__/",
+    ".mypy_cache/",
+    ".pytest_cache/",
+    ".tox/",
+    "dist/",
+    "build/",
+    ".eggs/",
+    "*.egg-info/",
+)
+
+_NOISE_SUFFIXES = (".pyc", ".pyo", ".pyd")
+
+
+def _is_noise(path: str) -> bool:
+    for prefix in _NOISE_PREFIXES:
+        if path.startswith(prefix) or ("/" + prefix) in ("/" + path):
+            return True
+    for suffix in _NOISE_SUFFIXES:
+        if path.endswith(suffix):
+            return True
+    return False
+
 
 def _is_first_commit() -> bool:
     result = subprocess.run(
@@ -34,7 +66,7 @@ def get_changed_files() -> list[str]:
     raw = result.stdout.strip()
     if not raw:
         return []
-    return [f for f in raw.split("\n") if f.strip()]
+    return [f for f in raw.split("\n") if f.strip() and not _is_noise(f)]
 
 
 def get_commit_message() -> str: