feat: add Atlas Cloud provider

[binyangzhu000-sudo]

Description

Add Atlas Cloud as a first-class, optional OpenAI-compatible provider. Existing provider defaults and routing behavior remain unchanged unless atlas is selected explicitly.

Type of Change

• Bug fix
• New feature
• Documentation update
• Performance improvement
• Refactoring

Changes Made

• add atlas configuration, validation, hot reload, tier/fallback routing, provider discovery, and lynkr init support
• translate Anthropic messages and tools to Atlas Chat Completions, including cross-format SSE transformation and reasoning-content handling
• default to the currently listed qwen/qwen3.8-max model and https://api.atlascloud.ai/v1/chat/completions, while allowing both to be overridden
• prevent automatic replay of billable Atlas Chat Completions POST requests
• document static and tier-based setup, and add an optional Atlas integration link with the official theme-aware logo
• add offline contract coverage for defaults, validation, request conversion, static routing, SSE registration, and no-retry behavior

Testing

• Added new tests
• Manual testing performed
• node --test test/atlas-integration.test.js test/dispatch-registry.test.js test/init.test.js (22 passed)
• npm run test:performance
• npm pack --dry-run
• live invokeAtlas smoke with qwen/qwen3.8-max returned HTTP 200 and the expected report_status({"status":"LYNKR_ATLAS_OK"}) tool call
• npm test is fully green: 1,256 passed and 1 unrelated TencentDB default assertion failed; the same failure reproduces on unmodified origin/main
• npm run lint is fully green: both this branch and unmodified origin/main report the same 110 pre-existing errors; this change adds none

Checklist

• Code follows the existing provider patterns
• Documentation updated
• Tests added/updated
• No secrets, generated artifacts, or local absolute paths included
• Commit includes a DCO sign-off