fix(telegram): route conflict-retry start_polling failure into network reconnect ladder

[shanewas]

Summary

Fixes #25221 — Telegram gateway stays alive with dead polling after the conflict-retry start_polling call fails.

The Bug

In _handle_polling_conflict(), when the retry start_polling call (line 896) raises an exception, the handler only logged the error and returned silently — no reconnect was scheduled, no further retry was triggered. This left the gateway process alive but with dead Telegram polling, requiring a manual restart.

The Fix

The fix mirrors what _handle_polling_network_error already does for network failures (lines 814-823): when start_polling fails in the conflict-retry path, schedule a background task that calls _handle_polling_network_error(retry_err). This routes the failure through the exponential-backoff reconnect ladder so the adapter recovers automatically instead of going silent.

Changed file: gateway/platforms/telegram.py (lines 904-916)

except Exception as retry_err:
    logger.warning("[%s] Telegram polling retry failed: %s", self.name, retry_err)
    # start_polling failed — polling is dead and no further error
    # callbacks will fire, so schedule the network reconnect handler
    # ourselves, routing the retry failure through the exponential-
    # backoff reconnect ladder so the adapter doesn't go silent.
    if not self.has_fatal_error:
        task = asyncio.ensure_future(
            self._handle_polling_network_error(retry_err)
        )
        self._background_tasks.add(task)
        task.add_done_callback(self._background_tasks.discard)
    return

Tests

tests/gateway/test_telegram_polling_conflict_retry.py — 4 regression tests:

• test_conflict_retry_start_polling_failure_schedules_network_reconnect — core fix validation
• test_conflict_retry_start_polling_failure_does_not_set_fatal_immediately — confirms transient path
• test_conflict_retry_start_polling_failure_leaves_adapter_alive — adapter stays running
• test_successful_conflict_retry_resets_conflict_count — conflict count resets on success

All 4 pass. Run: pytest tests/gateway/test_telegram_polling_conflict_retry.py -v

[alt-glitch]

Duplicate of #25232. Both PRs apply the identical fix to gateway/platforms/telegram.py (route start_polling() failure in _handle_polling_conflict() into _handle_polling_network_error via asyncio.ensure_future). #25232 was opened first. This PR adds more test coverage in a separate test file which could be cherry-picked.

[shanewas]

Closing as duplicate of #25232 (opened first, same fix). Will cherry-pick the test coverage from this PR into #25232.