feat: reject feedback prompt (60s window) + timeout UX cleanup

[wangyuyan-agent]

Summary

Two UX improvements for the Telegram provider:

1. Reject feedback prompt (60-second window)

After clicking Reject, the bot now sends a follow-up message:

❌ Rejected. Reply to this message within 60 seconds to add a reason (or ignore to skip).

• If the user replies within 60 s → feedback is captured in the JSON output ("feedback": "<reason>")
• If no reply arrives → exits normally with "feedback": null
• No change to exit codes; Reject is still exit code 1

This fills the pre-existing feedback field that was always null for rejected requests.

2. Timeout UX cleanup

Previously, when a request timed out the inline keyboard stayed on the original message with no indication that the request had expired.

Now on timeout:

• The inline keyboard is removed from the original message (editMessageReplyMarkup)
• A localised notice is sent: ⏰ Request timed out — no response received.
• Exit code 2 is unchanged

Localisation

New strings added for en, zh-CN, and zh-TW:

• reject_feedback_prompt
• reject_feedback_callback
• timeout_notice

Testing

cargo build  # passes with zero warnings

Manual flow:

1. openfeedback send --title "Test" --body "body" → click Reject → reply with reason → JSON shows "feedback": "your reason"
2. Same but ignore the follow-up → JSON shows "feedback": null
3. Let a short-timeout request expire → buttons removed, timeout notice appears

[antx-code]

Thanks for the PR! Both ideas are solid:

1. Reject feedback prompt — great call, filling the feedback field on reject is exactly what agents need
2. Timeout cleanup — removing stale buttons + sending a notice is the right UX

We're going to implement these ourselves with a few adjustments:

• Make the feedback wait window configurable (not hardcoded 60s)
• Have the follow-up message be a reply_to the original request, so the user replies to the correct message
• Keep the \u{xxxx} style for emoji consistency

Closing in favor of our own implementation. Thanks again for the contribution and the clean code — it made the review easy. 🙏