fix(copilot): claude-sonnet-4.6 max reasoning effort not available; model key mismatch on refresh #33607

Bug 1: max reasoning effort missing for claude-sonnet-4.6

The Copilot API returns reasoning_effort: ["low", "medium", "high", "max"] for claude-sonnet-4.6, but OpenCode only exposes low/medium/high.

Root cause in transform.ts:

• @ai-sdk/anthropic + github-copilot fallback path had efforts.filter(v => v !== "max" && v !== "xhigh") with stale comment "Efforts currently supported: low, medium, high"
• @ai-sdk/github-copilot claude fallback path used WIDELY_SUPPORTED_EFFORTS which excludes max

Verified: output_config.effort=max is accepted by /v1/messages and controls thinking depth; reasoning_effort=max accepted by /chat/completions.

Bug 2: model key mismatch causes metadata loss on every refresh

Static catalog keys claude models as claude-sonnet-4-6 (hyphen). Copilot API returns id: "claude-sonnet-4.6" (dot). The lookup remote.get(model.api.id) fails, so the catalog entry is deleted and re-added as a new entry on every refresh — without prev, losing any metadata carried from the catalog and breaking saved model preferences.