fix: defer Telegram streaming overflow chunks

[RichardAtCT]

Summary

• Avoid emitting Telegram overflow continuation messages during in-progress streaming edits
• Freeze the current preview once it exceeds Telegram's edit limit and let finalization deliver the missing tail once
• Add regression coverage that mid-stream overflow does not call edit/send continuation APIs

Tests

• uv run --extra dev python -m pytest tests/gateway/test_telegram_overflow_partial.py -q
• uv run --extra dev python -m pytest tests/gateway/test_stream_consumer.py tests/gateway/test_stream_consumer_fresh_final.py tests/gateway/test_duplicate_reply_suppression.py tests/gateway/test_telegram_overflow_partial.py -q

[alt-glitch]

This was generated by AI during triage.

Related (Telegram >4096-char streamed-overflow cluster): #50965 (open, flood-control cooldown + RetryAfter parsing), #48648 (open issue, infinite re-split loop), #42766 (closed, partial-overflow recovery). This PR addresses a distinct facet — deferring overflow continuation during in-progress streaming edits and delivering the tail once at finalization — so it is related, not a duplicate. Flagging the cluster for a maintainer to pick the consolidation path.

[teknium1]

Closing as superseded by #51736 (merged to main in 73a20a6ad), which fixed #48648 by gating the overflow split on finalize. Your fix targeted gateway/platforms/telegram.py, which no longer exists — the Telegram adapter was relocated to plugins/platforms/telegram/adapter.py in a bundled-plugin refactor, so the change landed there instead. The salvaged PR (#50408 by @Tranquil-Flow) hit the current path. Thanks for the fix and the report — credited in the issue.