Align DeepSeek and GLM execution with DS4

src/engine/metal/qwen_mtplx/request.rs

@@ -357,8 +357,8 @@ fn qwen_request_progress_uses_ui_frontiers_and_decode_timer() {
     let timing = PromptTiming {
         evaluated_tokens: 20,
         eval_seconds: 0.125,
-        mtp_history_seconds: 0.025,
-        restore_seconds: 0.010,
+        mtp_history_seconds: Some(0.025),
+        restore_seconds: Some(0.010),
     };
     assert_eq!(metrics.snapshot().prompt_timing, None);
     metrics.set_prompt_timing(Some(timing));
@@ -399,7 +399,7 @@ fn qwen_request_progress_uses_ui_frontiers_and_decode_timer() {
     let exact_hit = PromptTiming {
         evaluated_tokens: 0,
         eval_seconds: 0.0,
-        mtp_history_seconds: 0.0,
+        mtp_history_seconds: Some(0.0),
         ..timing
     };
     metrics.set_prompt_timing(Some(exact_hit));

src/engine/metal/qwen_mtplx/session_cache.rs

@@ -1163,8 +1163,8 @@ impl Execution {
         observe(TurnProgress::PromptReady(crate::metrics::PromptTiming {
             evaluated_tokens: prompt.len() - cached,
             eval_seconds: prepared.eval_seconds + history_seconds,
-            mtp_history_seconds: history_seconds,
-            restore_seconds,
+            mtp_history_seconds: Some(history_seconds),
+            restore_seconds: Some(restore_seconds),
         }))?;
         // The bank owns this boundary before decode starts, including when a
         // subsequent callback aborts the request. Keeping it in this stack