Better drawing and ordering of cyclic schedules with latency offsets

Some flaws (and probably a Matplotlib bug as the png does not clip in a similar way as the svg), but typically a nicer result.

Outputs are sorted to be directly below the operation that generates them. Dontcares are sorted to be directly above.

Closes #330 (closed)

b_asic/schedule.py

@@ -1194,7 +1194,7 @@ class Schedule:
             xy = np.stack((xvalues, y + y_pos))
             p = ax.add_patch(Polygon(xy.T, fc=_LATENCY_COLOR))
             p.set_clip_box(TransformedBbox(Bbox([[0, 0], [1, 1]]), ax.transAxes))
-            if any(xvalues > self.schedule_time):
+            if any(xvalues > self.schedule_time) and not isinstance(operation, Output):
                 xy = np.stack((xvalues - self.schedule_time, y + y_pos))
                 p = ax.add_patch(Polygon(xy.T, fc=_LATENCY_COLOR))
                 p.set_clip_box(TransformedBbox(Bbox([[0, 0], [1, 1]]), ax.transAxes))
@@ -1223,7 +1223,9 @@ class Schedule:
                     color=_EXECUTION_TIME_COLOR,
                     linewidth=3,
                 )
-                if any(xvalues > self.schedule_time):
+                if any(xvalues > self.schedule_time) and not isinstance(
+                    operation, Output
+                ):
                     ax.plot(
                         xvalues - self.schedule_time,
                         y + y_pos,