Infer worker types for releases, improve report formatting and diagnostics

[mouxinqq]

Motivation

• Reduce spurious unknown classifications for release / token release log lines by inferring worker type from nearby select events and make reports more informative.
• Surface optional config read errors (register.yaml) in error analysis so users can ignore them if not using the file.
• Improve table rendering and shorten overly long status reason lines to make generated troubleshoot reports easier to read.

Description

• Add _infer_release_worker_type and _infer_token_release_worker_type to scripts/log_parser.py and use them in match_select_release to infer release types from nearby selects, treating token-release log text as not fully trustworthy.
• Change match_select_release to store inferred release types and use inferred types when computing type_summary so prefill/mixed token counts are more accurate.
• Change token-release parsing to record raw_token_type and mark token releases as unknown_tokens until inferred.
• Add a focused self-test for token-release inference to log_parser.py's self-test flow.
• Add Failed to read YAML file config/register.yaml to SOURCE_LAYER_RULES in scripts/analyzers/errors.py and include an informational note in the errors report when that message appears.
• Limit the number of diagnoses shown in the concise load summary in scripts/analyzers/load_report.py, add explanatory notes about stats sampling and token-release attribution, and append an overflow indicator when many diagnoses exist.
• Improve Markdown table alignment generation in scripts/chart.py by emitting proper separator tokens for left/right alignment.
• Deduplicate and cap the reasons list in scripts/troubleshoot.py when determining DEGRADED status and add an explicit status definition line to the top of the full report.
• Update documentation references: error_catalog.md, log_patterns.md, and report_templates.md to reflect the new message, log pattern details, and status definitions.

Testing

• Ran log_parser.py self-test which includes the new token-release inference checks, and all self-test assertions passed.
• Performed local smoke formatting of load and error reports using the updated analyzer functions to verify the new summary truncation and informational messages render without error.
• Verified that render_table produces Markdown table separators with correct alignment markers by rendering sample tables locally.

---

Codex Task