feat: add Atlas Cloud provider preset

[binyangzhu000-sudo]

Summary

Adds Atlas Cloud as a built-in aggregator preset, shaped like the Requesty entry (not marked recommended).

Disclosure: I work at Atlas Cloud.

• src/shared/domain/presets.ts: new atlascloud preset after api-route, with three endpoints: OpenAI Chat (https://api.atlascloud.ai/v1, model list public), OpenAI Responses (same base), and Anthropic (https://api.atlascloud.ai, the client appends /v1/messages). Suggested models: deepseek-ai/deepseek-v4-flash, moonshotai/kimi-k3, zai-org/glm-5.2.
• Brand icon: lobehub ships an atlascloud glyph, so the brand is registered (BRANDS, rule in brands.ts, import in ProviderIcon.tsx) instead of falling back to initials.
• Tests: pinned Anthropic URL row in presets.test.ts, the preset added to the public model-list set, the brand table in preset-brands.test.ts, and a resolveBrand sample. baseurl.ts only gets the doc-table row.

verification is documented, not probed: the file's probed rule asks for a 401 on the real path and a 404 on a fake path under the same prefix, but Atlas Cloud answers 401 for any path under /v1 (the auth layer runs before routing). I checked the paths with real requests instead (below). Happy to change the level if you prefer.

The note on the card says that a successful model listing does not prove the key is valid: GET /v1/models is public and answers 200 even with a bad key.

Test plan

• vitest run on src/shared/domain, src/renderer/src/components/brand and src/renderer/src/settings/pages/model: all pass (52 files). On the whole suite the only failures are src/main/environment/__tests__/sftp.test.ts (also fails on a clean main) and 5 tests in broker-terminal / provider-file that fail only under full-suite load and pass when run alone.
• Live check through the app's own code (findPreset / endpointFor / previewUrl / modelListRequest / parseModelList / modelListAvailability) against the real API: the model list returns 200 with 119 models and no key needed; POST /v1/chat/completions, /v1/responses and /v1/messages (with x-api-key) each return a reply for deepseek-ai/deepseek-v4-flash.
• Not run: the Electron app UI (the preset card and icon were not looked at on screen), and tsc for the web/node projects: both fail on an unrelated existing test file (TS6307) on a clean main as well.

🤖 Generated with Claude Code