troubleshoot: enrich log analysis and reporting (select/release inference, cache cross-diagnosis, detailed outputs)

[mouxinqq]

Motivation

• Improve accuracy of select/release matching and worker-type attribution so counter/token metrics and unmatched events are less noisy.
• Provide richer cache/load/health/trace diagnostics and exportable detail files for deeper offline investigation.
• Make reports clearer about metric semantics and surface useful quick-impact info (e.g. YAML missing, detection latency).

Description

• Enhanced log_parser.match_select_release to infer release types and token release ownership, infer worker dominant types, expose unmatched_releases, and include richer id_coverage, type_summary, and worker_type_profile outputs, plus unit sample in self-test to validate token release inference.
• Added release/type inference helpers _infer_release_worker_type and _infer_token_release_worker_type in log_parser.py and improved matching logic to reduce unknown type counts.
• Extended load analyzers (load.py, load_report.py) to parse request-count select/release (count:), compute counter last-state (_analyze_counter_last_state), include counter_last_state in detail reports, and surface worker-type profiles and unmatched release lists in summaries and detail files.
• Added cross-diagnosis in cache analyzer (cache.py) via _analyze_cross_diagnosis and integrated detailed detail/cache_diagnosis.md output and links from the summary.
• Improved errors analyzer (errors.py) with source-impact mapping (IMPACT_RULES), classify_impact, and URL extraction for error templates, and surfaced Failed to read YAML file config/register.yaml as a special low-impact note.
• Health analyzer (health.py) now collects register events into timelines and computes detection latency _compute_detect_latency, adding a "detection latency" column to the worker table and detail output.
• Trace analyzer (trace.py) now runs match_select_release for each trace and includes select/release checks in trace diagnoses and detail data.
• Latency and errors reports annotated with metric definitions and link pointers to detailed diagnosis files.
• Chart/table rendering tweak in chart.py to adjust Markdown table separator alignment for right-aligned columns.
• Central report assembly (troubleshoot.py) expanded to collect and write multiple detail files under detail/ (e.g. load_select_release.md, load_counter_state.md, cache_* files, errors_topn.md, latency_diagnoses.md, and trace files) and to deduplicate DEGRADED reasons and add status definitions.
• Documentation snippets updated (error_catalog.md, log_patterns.md, report_templates.md) to reflect new messages, select/release log formats, and detail file layout.

Testing

• Ran log_parser.py self-test which includes the new token-release/type inference sample and basic parse/normalize checks, and it passed locally.
• Performed a quick smoke run of scripts/troubleshoot.py on sample logs to generate the summary and detail/ files (errors, load, cache, trace), and verified the new detail files are produced and referenced from the summary.
• No failing automated tests observed during the above smoke/self-test steps.

---

Codex Task