fix(channel): populate InboundMessage.mentioned_bot from mentions (fixes #134)

[Xuxchloris]

Fixes #134

Problem

InboundMessage.mentioned_bot is unconditionally False for every IM message, even when the bot is explicitly mentioned and the bot identity is resolved. Two compounding omissions in lark_oapi/channel/normalize/pipeline.py:

1. extract_mentions(raw_mentions) never receives bot_open_id, so mentions.py can never flag the bot self-mention.
2. Even when computed, ext.mentioned_bot is never forwarded to the InboundMessage(...) constructor — it falls back to its False default.

The internal PolicyGate works around this by re-computing the check inline (policy_gate.py), which confirms the field is known to be unreliable — but every on("message") consumer reading msg.mentioned_bot always gets False, causing "bot is @-mentioned but does not respond" bugs.

The comment path already does this correctly: _handle_comment_event passes bot_open_id=self._bot_open_id into normalize_comment.

Changes

• pipeline.py: process() accepts an optional bot_open_id; it is forwarded to extract_mentions, and ext.mentioned_bot is now passed to InboundMessage(...).
• channel.py: _handle_message_event passes bot_open_id=self._bot_open_id (same source the comment path uses).
• tests/test_pipeline.py: 3 new tests — bot in mentions[] → mentioned_bot=True (and the bot self-mention excluded from the public mentions list); no bot_open_id → False; bot not mentioned → False.

Verification

• python -m pytest lark_oapi/channel/tests — 660 passed; the single failure (test_upload_error_propagation.py::test_gather_buffer_missing_local_file_raises_upload_failed) is a pre-existing Windows-only path-escaping assertion that does not touch this code path (CI runs on Linux).

[CLAassistant]

All committers have signed the CLA.