Add Atlas Cloud LLM provider

[binyangzhu000-sudo]

Summary

• add Atlas Cloud as an optional OpenAI-compatible LLM provider
• expose API key and endpoint settings, model import support, and documentation
• preserve existing defaults while allowing credentials through settings or environment variables

Validation

• 27 passed across the focused provider, model core, and data configuration tests
• live API smoke test with openai/gpt-4.1-mini returned the expected response
• git diff --check

Atlas Cloud is opt-in and does not change the default provider.

[binyangzhu000-sudo]

Rebuilt on top of master (2.8.13) — conflicts resolved, the PR is mergeable again.

The only real conflict was docs/source/configuration.rst: the API keys chapter was restructured so each provider now gets its own heading instead of inline bold text, and Custom providers moved out into its own chapter. I rewrote the Atlas Cloud entry in the new style, placed it after Eden AI, and matched the wording pattern the neighbouring entries use. Upstream's new Layout section is kept as written.

One small tidy-up while I was in there: api_endpoint_atlas_cloud and api_key_atlas_cloud were sitting after the Eden AI keys in config.json, but that file keeps those two key families in alphabetical order — they're now between anthropic and deepseek.

I deliberately did not add a patch.py migration entry. updater.post_check_config() already backfills any base key missing from an existing profile, and it runs on the same version-mismatch path as the explicit to_add lists, so existing installs pick up the two new keys without one. Happy to add an entry under a 2.8.14 block if you'd rather have it explicit.

Verification on the merged branch:

• tests/provider/llms/test_llm_atlas_cloud.py + test_llm_custom.py → 9 passed
• docs/source/configuration.rst parses with docutils, no warnings (heading underline lengths check out)
• config.json / settings.json still parse, and the Atlas Cloud entries are intact in settings.json, locale.en.ini, core/types/openai.py, core/models/models.py, controller/model/importer.py and the launcher.add_llm(...) list in app.py
• AtlasCloudLLM.llama / get_models still match the current BaseLLM signatures