fix(qqbot): add return_exceptions=True to asyncio.gather in chunked upload

[x7peeps]

Summary

Add return_exceptions=True to the asyncio.gather() call in _run_with_concurrency used by the QQBot chunked file upload pipeline.

Problem

The asyncio.gather() at gateway/platforms/qqbot/chunked_upload.py:602 had no return_exceptions=True. When uploading a multi-part file, if any single part upload failed (network error, server rejection, etc.), gather() would cancel all other concurrent uploads rather than letting them finish, wasting bandwidth and potentially leaving the upload in an inconsistent state. This is the same pattern already fixed in trajectory_compressor (PR #65932), context_references (PR #62902), and feishu (PR #64864).

Fix

• Pass return_exceptions=True to asyncio.gather() so each part upload runs to completion regardless of sibling failures.
• Collect any exceptions from results and re-raise the first one, preserving the existing -> None return contract of _run_with_concurrency.

Testing

• Syntax check passed (py_compile)
• Behavior verified

[alt-glitch]

This was generated by AI during triage.

Related to #61302 at the same QQBot gather site. #61302's one-line return_exceptions=True change would swallow failures; this PR instead re-raises after sibling uploads complete. Maintainer choice is needed.

[x7peeps]

Closing as duplicate