Add optimizer issue handling experience to TiDB query tuning

[hawkingrei]

Summary

This PR adds optimizer issue-handling experience to the tidb-query-tuning skill.

It expands the skill from pure topic references into a mixed knowledge base that now includes:

• curated stats tuning references
• redacted optimizer oncall experience files
• generated GitHub issue-derived customer planner issue files
• agent guidance for mining new experience from GitHub issues
• a reusable script for generating issue-based experience corpora

What Changed

• added three new stats-focused reference docs for:
  • statistics health and auto analyze
  • statistics loading and startup behavior
  • stats version and analyze configuration
• copied the redacted optimizer oncall experience corpus into skills/tidb-query-tuning/references/optimizer-oncall-experiences-redacted/
• moved the generated customer planner issue corpus into skills/tidb-query-tuning/references/tidb-customer-planner-issues/
  • 175 issue files
  • 1 generated index README
• updated skills/tidb-query-tuning/SKILL.md so the skill points to topic docs, redacted field cases, and issue-derived precedents
• added skills/tidb-query-tuning/AGENTS.md with guidance on when and how to mine GitHub issue experience
• added skills/tidb-query-tuning/scripts/generate_tidb_issue_experiences.py to generate one-issue-one-file markdown corpora from GitHub issue and PR history

Why

General tuning guidance is not enough for many optimizer problems. A lot of useful handling knowledge exists in field incidents and customer-reported issues:

• user-facing symptoms
• investigation signals
• temporary workarounds
• fixed versions
• linked PRs and merge timestamps
• open gaps that still need attention

This PR makes that knowledge available from the skill itself instead of leaving it in ad hoc output directories.

Validation

• ran python3 skills/tidb-query-tuning/scripts/generate_tidb_issue_experiences.py --help
• ran python3 -m py_compile skills/tidb-query-tuning/scripts/generate_tidb_issue_experiences.py

Notes

• outputs/ was intentionally left out of the tracked changes. The generated issue corpus now lives under the skill's references/ directory.
• the copied oncall corpus is redacted and intended as reusable operational guidance, not as source-of-record incident material.