fix(telegram): reset send pool after pool timeouts

[yungchentang]

Summary

• reset PTB's general Bot API request pool after a confirmed Telegram pool timeout
• keep polling request pool resets separate so getUpdates recovery behavior stays unchanged
• add a regression test that pool timeouts drain _request[1] before retrying and leave _request[0] untouched

Fixes #53524

Tests

• python -m pytest tests/gateway/test_telegram_thread_fallback.py -q
• python -m pytest tests/gateway/test_telegram_network_reconnect.py -q
• python -m pytest tests/gateway/test_telegram_closewait_limits_31599.py -q
• python -m pytest tests/gateway/test_telegram_send_path_health.py -q
• python -m py_compile plugins/platforms/telegram/adapter.py tests/gateway/test_telegram_thread_fallback.py
• git diff --check

[alt-glitch]

This was generated by AI during triage.

Competing-PR pair with #53624 — both fix #53524 via the same mechanism (drain PTB's general request pool _request[1] after a send-side pool timeout). This PR is the earlier of the two (canonical); #53624 marked duplicate. Related cluster: #38812, #43333, #35624, #41391 (same #31599 pool-exhaustion family; root CLOSE_WAIT fix merged in #51541). Maintainer chooses which to merge.

[teknium1]

Merged via PR #54121 (#54121). Your commit was cherry-picked onto current main with your authorship preserved in git log (merge commit dddaea0). Thanks for the clean fix and the asyncio.Lock guard against concurrent drains.