driver: optimize the step_multiple step count calculation

this was using a stupid amount of compute.
we still have about 7% time unaccounted for. down from 12-15%.

crates/coremem/src/render.rs

@@ -535,6 +535,14 @@ impl<S> MultiRendererElement<S> {
             Some(end) => frame < end,
         }
     }
+    fn next_frame_for_work(&self, after: u64) -> Option<u64> {
+        let max_frame = after + self.step_frequency;
+        let max_frame = max_frame - max_frame % self.step_frequency;
+        match self.step_limit {
+            None => Some(max_frame),
+            Some(end) => Some(max_frame).filter(|&f| f < end)
+        }
+    }
 }
 
 pub struct MultiRenderer<S> {
@@ -567,6 +575,9 @@ impl<S> MultiRenderer<S> {
     pub fn any_work_for_frame(&self, frame: u64) -> bool {
         self.renderers.read().unwrap().iter().any(|m| m.work_this_frame(frame))
     }
+    pub fn next_frame_for_work(&self, after: u64) -> Option<u64> {
+        self.renderers.read().unwrap().iter().flat_map(|m| m.next_frame_for_work(after)).min()
+    }
 }
 
 impl<S: AbstractSim> Renderer<S> for MultiRenderer<S> {