fix: retry loop every 30s after panic, auto-reconnect on network recovery

fix: catch_unwind for telegram dispatcher panic (network timeout)

fix: add tracing subscriber in Telegram block for teloxide debug output

fix: Telegram block back to TOP, cfg!() macro, before any init_logging

fix: use cfg!() instead of #[cfg] block (compat issue)

fix: move telegram block after init_logging (tracing subscriber now active)

fix: telegram block at top of main(), #[cfg] no-label block, wait PID fix

nuclear: clean target dir, force full rebuild with latest source

debug: [TGRAM] markers to trace telegram flow

fix: keep process alive in bot error case, TcpListener stays up

fix: don't exit(1) on missing token, sleep to keep process alive

fix: telegram mode always on when feature enabled, clear error if token missing

fix: fallback to manual env var read for token (clap env feature unreliable)

debug: add eprintln tracing to see exactly where binary stops

Telegram Bot完整实现 + HF Space部署