Add Kimi (Moonshot AI) provider support

fix(thinking): ensure includeThoughts is false for ModeNone in budget processing

fix(thinking): support user-defined includeThoughts setting with camelCase and snake_case variants

fix(gemini): support snake_case thinking config fields from Python SDK

feat(thinking): enable thinking toggle for qwen3 and deepseek models

feat(registry): support provider-specific model info lookup

feat(claude): clamp thinking budget to max_tokens constraints

refactor(claude): move max_tokens constraint enforcement to Apply method

fix(thinking): update ValidateConfig to include fromSuffix parameter and adjust budget validation logic

refactor(thinking): add Gemini family provider grouping for strict validation

refactor(thinking): pass source and target formats to ApplyThinking for cross-format validation

feat(iflow): add iflow-rome model definition

refactor(thinking): extract antigravity logic into a dedicated provider

feat(thinking): support zero as a valid thinking budget for capable models

refactor(thinking): simplify ModeNone and budget validation logic

refactor(thinking): improve logging for user-defined models

fix(thinking): refine ModeNone handling based on provider capabilities

refactor(thinking): improve budget clamping and logging with provider/model context

feat(logging): order log fields for improved readability

refactor(thinking): refine configuration logging

fix(thinking): map reasoning_effort to thinkingConfig

test(thinking): remove legacy unit and integration tests

fix(thinking): improve model lookup and validation

fix(thinking): use static lookup to avoid alias issues

fix(executor): properly handle thinking application errors

fix(thinking): fix auth matching for thinking suffix and json field conflicts

refactor: improve thinking logic