Fix Total CPU % on /workers tab to normalize by total nthreads

The WorkerTable's Total CPU % divided by len(workers) instead of
sum(nthreads), producing incorrect values when workers have multiple
threads (e.g., 400% instead of 50%). This aligns the "cpu" column
with the already-correct "cpu_fraction" column, matching the same
fix pattern as PR #3897 for memory_percent.

Closes #8490

Author: ernestprovo23

distributed/dashboard/components/scheduler.py

@@ -4308,7 +4308,7 @@ def update(self):
                     total_data = (
                         sum(ws.metrics["cpu"] for ws in self.scheduler.workers.values())
                         / 100
-                        / len(self.scheduler.workers.values())
+                        / sum(ws.nthreads for ws in self.scheduler.workers.values())
                     )
                 elif name == "cpu_fraction":
                     total_data = (

distributed/dashboard/tests/test_scheduler_bokeh.py

@@ -564,6 +564,14 @@ async def test_WorkerTable(c, s, a, b):
     assert all(nthreads)
     assert nthreads[0] == nthreads[1] + nthreads[2]
 
+    # Total CPU should be normalized by sum(nthreads), not len(workers)
+    cpu = wt.source.data["cpu"]
+    total_nthreads = sum(ws.nthreads for ws in s.workers.values())
+    expected_cpu_total = (
+        sum(ws.metrics["cpu"] for ws in s.workers.values()) / 100 / total_nthreads
+    )
+    assert cpu[0] == expected_cpu_total
+
 
 @gen_cluster(client=True)
 async def test_WorkerTable_custom_metrics(c, s, a, b):