Docs improvement: document agent-friendly retry behavior when recall is empty right after remember

[Derojuu]

The docs mention that recall can return no matching memories right after a remember because indexing may lag. In an agent workflow, this can look like the memory write failed unless the prompt includes retry guidance.

Expected improvement:

Add a short agent pattern to the MCP reference:
1. call memwal_remember or memwal_remember_bulk
2. wait briefly if immediate recall is empty
3. retry memwal_recall
4. only call memwal_restore when prior durable memories are expected but missing across sessions or after index loss

Evidence to attach after testing:

- Approximate delay observed:
- Number of retries needed:
- Namespace:
- Whether remember returned success:
- Whether later recall succeeded:

[github-actions]

Thank you for opening this issue, a team member will review it shortly.