Add Atlas Cloud LLM provider shortcut

[binyangzhu000-sudo]

Summary

• Add an Atlas Cloud LLM provider shortcut (atlascloud, atlas-cloud, atlas) that resolves to the existing OpenAI-compatible route.
• Default the shortcut to https://api.atlascloud.ai/v1, qwen/qwen3.5-flash, and ATLASCLOUD_API_KEY / ATLAS_CLOUD_API_KEY env var references.
• Wire the shortcut into compact summarization, maintenance task LLM routing, and memsearch config init model defaults.
• Add focused tests for config resolution, compact dispatch, and maintenance routing.

No README/docs/logo changes.

Validation

• uv run --python /Users/zby/.local/bin/python3.11 --group dev pytest tests/test_config.py tests/test_compact.py tests/test_maintenance.py (66 passed)
• uv run --python /Users/zby/.local/bin/python3.11 --group dev ruff check src/memsearch/config.py src/memsearch/compact.py src/memsearch/maintenance.py src/memsearch/cli.py tests/test_config.py tests/test_compact.py tests/test_maintenance.py
• PYTHONPYCACHEPREFIX=/tmp/memsearch_pycache python3 -m compileall src/memsearch/config.py src/memsearch/compact.py src/memsearch/maintenance.py src/memsearch/cli.py tests/test_config.py tests/test_compact.py tests/test_maintenance.py
• git diff --check
• Atlas Cloud live model catalog returned 436 models and confirmed qwen/qwen3.5-flash and deepseek-ai/deepseek-v4-pro.