fix(discord): reuse existing starter threads

[chenjian-agent]

Summary

Fix Discord message handling when a starter thread already exists for the triggering message.

This makes get_or_create_thread() resilient instead of treating Discord's existing-thread response as a fatal error.

Closes #189

Root cause

The previous implementation always called create_thread_from_message(...) for non-thread messages.

If Discord had already created a starter thread for that message, the API returned:

A thread has already been created for this message

OpenAB logged the error and returned early, so the prompt was dropped even though a valid thread already existed.

Changes

• import MessageFlags
• reuse msg.thread when Discord includes the starter thread on the message payload
• reuse msg.id when MessageFlags::HAS_THREAD is present
• treat Discord's A thread has already been created for this message response as a recoverable case and fall back to the existing starter thread id

Validation

• reproduced the production failure from logs on 2026-04-10
• built locally with cargo build
• deployed the patched binary on the production host
• restarted openab.service
• confirmed the service came back healthy and connected after restart

Why this approach

For message-based starter threads, Discord uses the starter message id as the thread id. That means the correct recovery path is to reuse the existing thread instead of failing the request.

[chenjian-agent]

Production debugging notes for this fix:

1. We first verified that the service itself was not down. openab.service remained active (running) and the bot was still connected to Discord.
2. We checked the runtime logs and found repeated failures of the form:
   • failed to create thread: A thread has already been created for this message
3. We traced the failure path to get_or_create_thread() in src/discord.rs.
4. The existing logic always attempted create_thread_from_message(...) for non-thread messages and returned early on that Discord error.
5. We confirmed that this made the bot drop otherwise valid requests whenever the triggering message already had a starter thread.
6. We patched the handler to reuse the existing thread instead of failing:
   • return the current channel id when already inside a thread
   • reuse msg.thread when Discord includes the starter thread on the message payload
   • reuse msg.id when MessageFlags::HAS_THREAD is present
   • treat A thread has already been created for this message as a recoverable case and fall back to the existing starter thread id
7. We validated the code with a local cargo build.
8. We then deployed the patched binary on the production host and restarted openab.service to restore service.

This PR is the upstreamed version of that production fix.