Add profiler warmup overhead tracking and tooltip display

Nerfed.Runtime/Profiler.cs

@@ -434,6 +434,7 @@ public static class Profiler
         public int ManagedThreadId { get; private set; }
         public List<ScopeNode> Children { get; } = new List<ScopeNode>();
         public long AllocatedBytes { get; private set; }
+        public long ProfilerSetupBytes { get; private set; }
         internal ScopeNode Parent { get; private set; }
         internal long ChildrenDurationTicks { get; private set; }
         internal long ChildrenAllocatedBytes { get; private set; }
@@ -451,6 +452,7 @@ public static class Profiler
             ChildrenDurationTicks = 0;
             ChildrenAllocatedBytes = 0;
             AllocatedBytes = 0;
+            ProfilerSetupBytes = 0;
             Children.Clear();
             allocatedBytesAtStart = GC.GetAllocatedBytesForCurrentThread();
         }
@@ -463,9 +465,11 @@ public static class Profiler
             }
 
             EndTime = Stopwatch.GetTimestamp();
-            AllocatedBytes = Math.Max(0, GC.GetAllocatedBytesForCurrentThread() - allocatedBytesAtStart);
             if (Parent != null)
             {
+                // Root nodes are ended from FinalizeCurrentFrame on the main thread, so their
+                // GC counter would be from the wrong thread. Only track alloc on non-root nodes.
+                AllocatedBytes = Math.Max(0, GC.GetAllocatedBytesForCurrentThread() - allocatedBytesAtStart);
                 Parent.ChildrenDurationTicks += Math.Max(0, EndTime - StartTime);
                 Parent.ChildrenAllocatedBytes += AllocatedBytes;
             }
@@ -488,6 +492,12 @@ public static class Profiler
             return Math.Max(0, AllocatedBytes - ChildrenAllocatedBytes);
         }
 
+        // Called after all profiler setup (Children.Add + scopes.Push) to measure overhead within this scope's window.
+        internal void NoteSetupOverhead()
+        {
+            ProfilerSetupBytes = Math.Max(0, GC.GetAllocatedBytesForCurrentThread() - allocatedBytesAtStart);
+        }
+
         internal ScopeNode AddChild(string label, string category, ulong tagMask)
         {
             ScopeNode child = RentNode(label, category, tagMask, ManagedThreadId, this);
@@ -648,6 +658,7 @@ public static class Profiler
 
         ScopeNode newScope = scopes.Peek().AddChild(label, category, tagMask);
         scopes.Push(newScope);
+        newScope.NoteSetupOverhead();
     }
 
     [Conditional("PROFILING")]