feed-memex: emit ai-title record so memex shows readable chat names

Memex's importer reads ai-title records from inbox JSONLs to give each
imported chat a human-readable title in list_conversations / search.
We had session.title in scope already (extracted by the scanner), but
weren't passing it through — so every Claude Code/Cowork chat in
memex appeared as 'Claude Code · <8-char-file-stem>'.

Now feed-memex prepends one ai-title line to each output JSONL.
Existing memex installations pick this up via upsertConversation's
ON CONFLICT update on the next feed.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: sedelev

claude_backup/cli.py

@@ -396,6 +396,18 @@ def feed_memex_cmd(ctx: click.Context, inbox: Path, dry_run: bool) -> None:
                 )
             else:
                 with target.open("w", encoding="utf-8") as f:
+                    # Memex looks for an `ai-title` record to give the chat
+                    # a human-readable name in list_conversations / search.
+                    # Without this it falls back to the first user message
+                    # (or the file stem if there is none).
+                    if session.title:
+                        f.write(
+                            json.dumps(
+                                {"type": "ai-title", "aiTitle": session.title},
+                                ensure_ascii=False,
+                            )
+                            + "\n"
+                        )
                     for line in lines:
                         f.write(json.dumps(line, ensure_ascii=False) + "\n")
                 written += 1