Summary

• Adds a Kanban retry-loop guard so repeated timeout/spawn-failure patterns stop respawning indefinitely
• Blocks/gives up after the configured threshold with human-visible event trail and metadata
• Resets failure counters on reassignment so legitimate retries are preserved

Verification

• git diff --check
• python -m py_compile hermes_cli/kanban_db.py hermes_cli/kanban.py gateway/run.py hermes_cli/config.py
• python -m pytest tests/hermes_cli/test_kanban_core_functionality.py -q -k 'spawn_failure_auto_blocks_after_limit or reassign_resets_failure_counter or repeated_timeouts_auto_block' -o 'addopts='

Note: Forge reported full tests/hermes_cli/test_kanban_core_functionality.py has unrelated existing state-isolation failures: test_detect_crashed_workers_reclaims, test_board_stats, test_cli_create_on_fresh_home_auto_inits.