fix(cli): default image input to true for OpenAI-compatible providers (#8729)

Models fetched via the OpenAI-compatible /v1/models endpoint had image
support hardcoded to false because the endpoint provides no capability
metadata. This matches the legacy extension behavior (supportsImages:
true) for custom providers while leaving all other provider defaults
untouched.

packages/opencode/src/provider/model-cache.ts

@@ -209,7 +209,7 @@ export namespace ModelCache {
         name: model.id,
         family: model.owned_by ?? "",
         release_date: "",
-        attachment: false,
+        attachment: true,
         reasoning: false,
         temperature: true,
         tool_call: true,
@@ -217,7 +217,7 @@ export namespace ModelCache {
         limit: { context: 128000, output: 4096 },
         options: {},
         modalities: {
-          input: ["text"],
+          input: ["text", "image"],
           output: ["text"],
         },
       }