feat: add Requesty provider preset

[Thibaultjaigu]

Summary

Adds Requesty (an OpenAI compatible LLM router) as a built in aggregator preset, shaped the same way as the OpenRouter entry. It is not marked as recommended.

Changes

• src/shared/domain/presets.ts: new requesty preset next to openrouter with three endpoints: OpenAI Chat and Responses at https://router.requesty.ai/v1, Anthropic at https://router.requesty.ai (the client adds /v1/messages). Model list is enabled and marked public, since GET /v1/models returns 200 without a key. Includes docs and API key links, and a note about the EU base URL (router.eu.requesty.ai).
• src/shared/domain/baseurl.ts: Requesty added to the /v1 row of the prefix table.
• src/shared/domain/__tests__/presets.test.ts: Anthropic URL row for requesty, and requesty added to the modelListPublic set.
• src/renderer/src/settings/pages/model/__tests__/preset-brands.test.ts: lobehub has no Requesty icon, so it falls back to initials like OhMyGPT. The expected table and the no icon list are updated.

How to test

1. Settings, Model, add provider, pick Requesty under aggregators.
2. Paste a key from https://app.requesty.ai/api-keys (or use REQUESTY_API_KEY in a script).
3. Fetch the model list, or add openai/gpt-4o-mini manually and send a message.

Validation

• npm run typecheck (node and web): pass
• eslint on the touched files: pass
• vitest for src/shared/domain, src/renderer/src/settings/pages/model, src/renderer/src/components/brand, src/main/kernel/upstream: 65 files, 1760 tests pass
• Full vitest run: the failures I saw (document-engine native host, provider-registry, sftp, plugin document-rpc and document-scope) fail the same way on a clean main checkout on my machine. release-notes.test.ts timed out once during a full run and passed on rerun
• Live check through UpstreamRouter built from the preset endpoints: openai-chat and openai-responses with openai/gpt-4o-mini, and anthropic with claude-fable-5.1 all streamed a reply to message_end. modelListRequest plus parseModelList returned 200 both without a key (717 models) and with a key.

Disclosure: I work at Requesty. Happy to adjust anything to match project conventions.