feat(ai,coding-agent): add Requesty as a built-in provider

[devin-ai-integration]

Summary

Adds requesty as a built-in provider in @earendil-works/pi-ai and wires it into the coding agent (REQUESTY_API_KEY, default model, docs). Prepared against current upstream badlogic/pi-mono main (upstream-main branch here) so it can be submitted upstream once a maintainer has granted lgtm per pi's CONTRIBUTING.md.

Model catalog is generated, not hand-written: fetchRequestyModels() in scripts/generate-models.ts reads https://router.requesty.ai/v1/models (no auth), keeps api === "chat" && supports_tool_calling, and maps:

anthropic/*  -> api: "anthropic-messages", baseUrl: "https://router.requesty.ai"     (native /v1/messages: thinking, cache_control pass through)
everything else -> api: "openai-completions", baseUrl: "https://router.requesty.ai/v1"
input_price/output_price/cached_price/caching_price ($/token) * 1e6 -> cost.{input,output,cacheRead,cacheWrite}
supports_role_developer === false -> compat.supportsDeveloperRole: false

Yields 660 models (11 Anthropic Messages, 649 OpenAI completions). Provider factory requestyProvider() uses envApiKeyAuth + the existing lazy anthropicMessagesApi/openAICompletionsApi; no new API implementation. Coding-agent default is anthropic/claude-sonnet-4-6.

Verification

• npm run check clean; test/providers.test.ts (new Requesty auth/routing test) and test/env-api-keys.test.ts pass.
• Live smoke test with a real key: anthropic/claude-sonnet-4-6, openai/gpt-5.4-mini, moonshot/kimi-k2.6 all stream, call tools (stopReason: toolUse), and report usage/cost through the router.

Notes

• packages/ai/src/providers/data/requesty.json is gitignored like the other provider JSON; CI regenerates it. requesty.models.ts and models.generated.ts are included as generated output.
• Model name is the id (Requesty's description field is a paragraph, not a display name).
• This supersedes the earlier reverted feat(ai,coding-agent): add Requesty as native provider #1 approach; it follows the current upstream provider architecture (per-provider factory + providers/all.ts) that did not exist then.

Link to Devin session: https://app.devin.ai/sessions/e7290f77ce644b3f849f89535915f2d8
Open in Devin Desktop: https://app.devin.ai/desktop/session/e7290f77ce644b3f849f89535915f2d8?variant=devin
Requested by: @requesty-JohnCosta27

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment, CI, and merge conflict monitoring