Nenya 0.8.2

Changelog

• 3d3e4b6 Merge phase-001-cached-tokens: Z.AI cached_tokens tracking
• fb9f807 Merge phase-002-message-sanitization: sanitization wiring fix
• 2abe59b Merge phase-003-thinking-config: GLM Thinking config
• 73730d4 Merge phase-004-provider-name-fix: provider name fix
• 7b025eb feat(config): add Thinking config to GLM models and correct context/output limits
• 95354dc feat(stream): support nested cached_tokens for Z.AI
• 9ff60f3 fix(providers): check both provider names in thinking injection
• 000b0e0 fix(routing): wire zaiSanitize into specs and fix tools-skip orphan filtering