feat: report failed runner pages

[cafitac]

Summary

• add pages_failed to runner progress outputs so harness-facing surfaces can distinguish failed pages from completed or merely attempted ones
• count failed pages for terminal request failures like non-success status and retry exhaustion in both sequential and concurrent runner paths
• propagate the new field through crawl results, test reports, and CLI summaries with coverage updates

Testing

• uv run --extra dev pytest tests/unit/core/models/test_models.py tests/unit/core/agent/test_recipe_testing.py tests/unit/core/agent/test_agent_recipe_flow.py tests/unit/cli/test_run_command.py tests/unit/cli/test_test_recipe_command.py tests/component/core/runner/test_recipe_runner.py -q
• bash scripts/check-python.sh
• bash scripts/verify-ai-harness.sh
• inline CLI smoke asserting pages_failed=0 in ai-crawler run output