feat: persist inbound quoted_message_id and add reply support to /api/send

[blazer5000]

Closes #107

What this adds

Inbound persistence — quoted-reply detection was already implemented in extractQuotedMessageInfo (used for webhook forwarding). This PR wires the extracted quotedMessageId into StoreMessage so replies are also archived in SQLite.

A new nullable quoted_message_id TEXT column is added via ensureColumn — the existing migration helper, so it's safe on all existing installs with no manual action needed.

Column	Value
quoted_message_id	StanzaID of the replied-to message (NULL for plain messages)

Plain messages store NULL; the COALESCE merge in ON CONFLICT means a later history-sync re-store (which passes "") cannot clobber a non-null ID already written by handleMessage.

Outbound quoted replies — POST /api/send now accepts three optional fields:

{
  "recipient": "[email protected]",
  "message": "Great point!",
  "quoted_message_id": "3AORIGINAL0000001",
  "quoted_sender_jid": "[email protected]",
  "quoted_content": "original text"
}

When quoted_message_id is non-empty and no media_path is provided, the bridge sends an ExtendedTextMessage with a ContextInfo block so the message renders as a quoted reply on the recipient's device. Limitation: text-only — quoting media messages is not supported because the quote preview requires the original media key/URL, which the API caller does not have.

Outbound rows also persist quoted_message_id so list_messages can report it.

MCP server — send_message() and the send_message MCP tool both accept quoted_message_id, quoted_sender_jid, and quoted_content (all optional, default ""). They are forwarded to /api/send only when non-empty. list_messages and get_message_context now include quoted_message_id in every returned message dict (null for non-replies).

Tests

Go (whatsapp-bridge/main_test.go):

• TestHandleMessage_QuotedReply_IDPersisted — inbound reply stores correct quoted_message_id
• TestHandleMessage_PlainMessage_QuotedIDIsNull — plain message stores SQL NULL
• TestSendHandler_QuotedReplyFields_PassedThrough — new JSON fields are parsed (empty recipient → 400)
• TestExtractQuotedMessageInfo_ExtendedText — helper extracts StanzaID, Participant, and content
• TestExtractQuotedMessageInfo_NoContextInfo — plain/nil message returns all-empty

Python (tests/test_bridge_auth.py, tests/test_whatsapp.py):

• test_send_message_with_quoted_reply_includes_quote_fields — payload contains all three fields
• test_send_message_without_quote_omits_quote_fields — no extra keys when not quoting
• test_msg_to_dict_quoted_message_id_present/absent

All existing tests continue to pass (go test ./... and uv run pytest -v).

Schema impact

One new nullable column: messages.quoted_message_id TEXT. Added via ensureColumn — idempotent, backward-compatible, no downtime.