Add "Ollama Cloud" as a provider option

[danmurf]

Motivation

Buzz's agent provider picker has named entries for Anthropic, OpenAI, OpenRouter, and Databricks, but nothing for Ollama Cloud. Users who want to run hosted open models (e.g. kimi-k3, glm-5.2, deepseek-v4-pro) can technically get there today by picking "OpenAI-compatible" and manually typing the base URL, API key, and model id — but nothing in the UI hints this is possible, so most don't realise it's an option.

Proposed solution

Add "Ollama Cloud" as a named provider in the picker. A user selects it, pastes their Ollama API key into the credential field, and chooses a model from the auto-populated list — the same one-click flow the other named providers already offer. No manual env-var editing.

Alternatives considered

• Keep using the existing "OpenAI-compatible" entry (works today but undiscoverable — users have to know the magic base URL and that it's even an option).
• Fold this into Share compute: route to external OpenAI-compatible endpoints — local servers (Ollama / LM Studio / MLX / vLLM) and, opt-in, cloud/enterprise gateways #3410 (share-compute routing to external endpoints) — different concern: that centralizes a credential on a community serving node; this is per-user provider selection in the desktop picker.

Additional context

Ollama Cloud exposes an OpenAI-compatible endpoint, so no new wire format is needed — this is a discoverability and UX gap, not a capability gap.

Related: #3145 (local Ollama, sibling), #3410 (share-compute external endpoints).

[Chessing234]

opened #3921 — named Ollama Cloud provider that defaults to https://ollama.com/v1 and reuses the openai-compat key field so you don't have to hunt for the magic base URL.