fix(notes): take last numeric segment of href, not the first

The notes listing renders each thread link as
/msg/pms/{folder}/{noteID}/#message. extractIntFromHref returned the
first numeric segment it found, which was always the folder index (1
for the inbox), so every NotePreview.ID came out as 1 and any
follow-up GetNote(np.ID) call failed with "this message has either
been deleted or is not yours".

Surfaced by an end-to-end smoke run against the live site. Limited
to the notes parser; the other extractIntFromHref callers
(/view/{id}/, /journal/{id}/) only ever have a single numeric segment
so they are unaffected.

notes_parser.go

@@ -52,11 +52,17 @@ func parseNoteListItem(item *goquery.Selection) *NotePreview {
 	}
 
 	// Note ID lives in the href: /msg/pms/{folder}/{id}/#message. Strip the
-	// fragment first so extractIntFromHref picks the trailing numeric path.
+	// fragment first, then take the *last* numeric segment — the folder
+	// number (e.g. 1) appears before the note ID and would otherwise win
+	// the "first numeric segment" race in extractIntFromHref.
 	if i := strings.Index(href, "#"); i != -1 {
 		href = href[:i]
 	}
-	np.ID = NoteID(extractIntFromHref(href))
+	for _, seg := range strings.Split(href, "/") {
+		if n, err := parseID[NoteID](seg); err == nil && n != 0 {
+			np.ID = n
+		}
+	}
 
 	// Read/unread: classes on the subject link.
 	if class, _ := subjectLink.Attr("class"); strings.Contains(class, "note-unread") || strings.Contains(class, "unread") && !strings.Contains(class, "note-read") {