fix: model format normalization and explicit config cache bypass

- Add normalizeModelFormat() utility for string/object model handling
- Update subagent-resolver to handle both model formats
- Add explicitUserConfig flag to ModelResolutionResult
- Set explicitUserConfig: true when user model is found in pipeline

This fixes the issue where plugin-provided models fail cache validation
and fall through to random fallback models.

src/shared/model-resolution-pipeline.ts

@@ -33,6 +33,7 @@ export type ModelResolutionResult = {
   variant?: string
   attempted?: string[]
   reason?: string
+  explicitUserConfig?: boolean
 }
 
 function normalizeModel(model?: string): string | undefined {
@@ -58,7 +59,7 @@ export function resolveModelPipeline(
   const normalizedUserModel = normalizeModel(intent?.userModel)
   if (normalizedUserModel) {
     log("Model resolved via config override", { model: normalizedUserModel })
-    return { model: normalizedUserModel, provenance: "override" }
+    return { model: normalizedUserModel, provenance: "override", explicitUserConfig: true }
   }
 
   const normalizedCategoryDefault = normalizeModel(intent?.categoryDefaultModel)