feat(config): add MiniMax as a native model provider

[octo-patch]

Reason: Add MiniMax as a first-class native model provider selectable in setup and provider switching.

What changed

• Register minimax in the provider option list so it shows up in the setup modal and the /provider switch UI (both iterate the option list, so no further UI wiring is needed).
• Add the provider's default model configuration: the global OpenAI-compatible base URL (https://api.minimax.io/v1, to which the agent appends /chat/completions), an entry-level flash model default and a flagship pro model default, and a 1M context window. The completion limit is left unset so each model's own output default applies, matching the existing entries that opt out of a hard cap.
• Document the China base URL (https://api.minimaxi.com/v1) inline; advanced users can select it by editing model.yaml or by configuring a custom provider, so their overrides stay intact.
• Extend the provider defaults table test with a case that pins the new base URL, flash/pro model ids, context window, and completion setting.

Checks

• gofmt -l config/config.go config/config_test.go (clean)
• go vet ./config/
• go test ./config/
• go build ./...