feat: add NeuronPool OpenAI-compatible provider + Add Model wizard

[dannymota]

Adds NeuronPool as an OpenAI-compatible provider for Continue.

Changes

• Docs page at docs/customize/model-providers/more/neuronpool.mdx
• Add Model wizard row in gui/src/pages/AddNewModel/configs/providers.ts (provider: openai, live Worker apiBase, AUTODETECT models)

Until api.neuronpool.dev resolves, recipes use the live Worker:

• Base: https://neuronpool.damnknee.workers.dev/v1
• Key: sk-neuronpool-… from https://neuronpool.damnknee.workers.dev/dashboard
• Default model: gpt-oss-20b (GET /v1/models for the catalog)

The wizard file is ~47.9KB. The NeuronPool row was inserted via a fork Action (commit 6a1e496835f5b3e16348c27f02469cc432d37d4f, blob 8870eb275dfb9690f29f27c3df30ea2d248242b6) so the rest of providers.ts was not rewritten.

Verification:

• GET /v1/models returns catalog ids with a live buyer key
• Streaming chat completion replies (Continue CLI against a local pool)
• Tool-call at the NeuronPool API: local POST /v1/responses + tool_choice function → 200 function_call + X-NeuronPool-Pool (2026-09-04). Continue CLI + llama-3.2-1b-instruct still hallucinates file contents instead of calling tools — that is the 1B model, not a missing wizard row.
• No key committed; env var / password field only

[github-actions]

All contributors have signed the CLA ✍️ ✅
_{Posted by the CLA Assistant Lite bot.}

[dannymota]

Wizard row is now on this PR (gui/src/pages/AddNewModel/configs/providers.ts, commit 6a1e496). Inserted next to openrouter via a fork Action so the ~45.7KB file was not MCP-truncated. Blob SHA 8870eb275dfb9690f29f27c3df30ea2d248242b6; file still ends with the Nous block.

[dannymota]

I have read the CLA Document and I hereby sign the CLA

[dannymota]

recheck