fix(vault): tolerate BOM and leading blank lines in frontmatter parsing

A note saved by Notepad/PowerShell carries a leading UTF-8 BOM (and
sometimes leading blank lines) before the opening `---`. The old
\A-anchored regex treated that as "no frontmatter", dumping the whole
YAML block into the body -- which then hits the destructive
no-valid-ULID path in index.py. parse() now strips a leading BOM and
tolerates leading whitespace before the delimiter.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0139vHnNpS2sLNELjrAHX5Cu

src/mdmem/notes.py

@@ -19,7 +19,7 @@ from ruamel.yaml.comments import CommentedMap
 from mdmem.config import NOTE_TYPES
 from mdmem.ids import filename_for, new_id, nfc
 
-_FM_RE = re.compile(r"\A---[ \t]*\r?\n(.*?)\r?\n---[ \t]*\r?\n?", re.DOTALL)
+_FM_RE = re.compile(r"\A\s*---[ \t]*\r?\n(.*?)\r?\n---[ \t]*\r?\n?", re.DOTALL)
 _WIKILINK_RE = re.compile(r"\[\[([^\[\]|#]+?)(?:#[^\[\]|]*)?(?:\|[^\[\]]*)?\]\]")
 _HEADING_RE = re.compile(r"^(#{1,6})\s+(.*?)\s*#*\s*$", re.MULTILINE)
 
@@ -62,7 +62,10 @@ def parse(text: str, *, strict: bool = False) -> tuple[CommentedMap, str]:
     strict=True re-raises broken YAML — used where a human just typed it. The
     scanner stays lenient so one bad file cannot stop the whole vault indexing.
     """
-    text = nfc(text.replace("\r\n", "\n"))
+    text = text.replace("\r\n", "\n")
+    if text.startswith(""):  # BOM left by Notepad/PowerShell
+        text = text[1:]
+    text = nfc(text)
     m = _FM_RE.match(text)
     if not m:
         return CommentedMap(), text

tests/test_notes.py

@@ -82,6 +82,20 @@ def test_missing_frontmatter_is_tolerated():
     assert body.strip() == "just a body"
 
 
+def test_parse_tolerates_leading_utf8_bom():
+    """A note saved by Notepad/PowerShell carries a leading BOM (finding 2)."""
+    fm, body = parse("" + SAMPLE)
+    assert fm["title"].startswith("llama-swap")
+    assert body.startswith("llama-swap routes models")
+    assert "---" not in body
+
+
+def test_parse_tolerates_leading_blank_lines_before_delimiter():
+    fm, body = parse("\n\n  \n" + SAMPLE)
+    assert fm["title"].startswith("llama-swap")
+    assert body.startswith("llama-swap routes models")
+
+
 def test_wikilinks_with_alias_and_heading():
     n = Note.new(title="T", body="see [[target|alias]] and [[other#Head]]", type="note")
     assert n.wikilinks == ["target", "other"]