Record the model that served a run's usage

Why: a fallback chain can answer with a different model than the one that was requested, and usage has to name the model that actually served.
Evidence: go test .
Impact: the field is omitted when empty, so existing usage JSON stays the same.
Refs: feature/durable-rewind
Co-authored-by: Cursor <cursoragent@cursor.com>

model.go

@@ -201,6 +201,8 @@ type Usage struct {
 	TotalTokens      int     `json:"totalTokens"`
 	DurationMs       float64 `json:"durationMs"`
 	ToolCalls        int     `json:"toolCalls"`
+	// Model 是实际给出这段用量的模型。空表示当时没有记录。
+	Model string `json:"model,omitempty"`
 }
 
 // ---- 工件 ----

ts/types.ts

@@ -54,6 +54,7 @@ export type RunPhase = 'thinking' | 'acting' | 'responding' | 'finalizing' | 'do
 export interface Usage {
   promptTokens: number; completionTokens: number; totalTokens: number;
   durationMs: number; toolCalls: number;
+  model?: string;
 }
 
 export interface Run {