TargetEntry optimizations for Results (#9590)

Context
@jeffkl has a case with potentially thousands of items that would all go into one bucket in the Returns. MSBuild is deduplicating it, but so is NuGet, which is wasteful, so he will likely stop using Returns. This change will save him a small amount of allocations, but it would save a ton of allocations if he were to keep using it, since it will no longer need to repeatedly reallocate the HashSet, and it won't need to allocate as many Lists. It will still save one List's worth of allocations on every single Target that doesn't have a Returns.

Changes Made
Effectively making List allocations lazier and taking advantage of ToList to avoid doing as many (duplicated) collection operations.

src/Build/BackEnd/Components/RequestBuilder/TargetEntry.cs

@@ -4,6 +4,7 @@
 using System;
 using System.Collections.Generic;
 using System.Diagnostics;
+using System.Linq;
 using System.Threading;
 using System.Threading.Tasks;
 using Microsoft.Build.Collections;
@@ -568,7 +569,7 @@ internal async Task ExecuteTarget(ITaskBuilder taskBuilder, BuildRequestEntry re
                 }
 
                 // Produce the final results.
-                List<TaskItem> targetOutputItems = new List<TaskItem>();
+                TaskItem[] targetOutputItems = Array.Empty<TaskItem>();
 
                 try
                 {
@@ -622,37 +623,50 @@ internal async Task ExecuteTarget(ITaskBuilder taskBuilder, BuildRequestEntry re
 
                         if (keepDupes)
                         {
+                            List<TaskItem> targetOutputItemsList = new();
                             foreach (ItemBucket bucket in batchingBuckets)
                             {
-                                targetOutputItems.AddRange(bucket.Expander.ExpandIntoTaskItemsLeaveEscaped(targetReturns, ExpanderOptions.ExpandAll, targetReturnsLocation));
+                                if (targetOutputItems is null)
+                                {
+                                    // As an optimization, use the results for the first bucket and if there are no more buckets to process, only a single list is allocated.
+                                    targetOutputItemsList = bucket.Expander.ExpandIntoTaskItemsLeaveEscaped(targetReturns, ExpanderOptions.ExpandAll, targetReturnsLocation).ToList();
+                                }
+                                else
+                                {
+                                    targetOutputItemsList.AddRange(bucket.Expander.ExpandIntoTaskItemsLeaveEscaped(targetReturns, ExpanderOptions.ExpandAll, targetReturnsLocation));
+                                }
                             }
+
+                            targetOutputItems = targetOutputItemsList.ToArray();
                         }
                         else
                         {
-                            HashSet<TaskItem> addedItems = new HashSet<TaskItem>();
-                            foreach (ItemBucket bucket in batchingBuckets)
+                            // Optimize for only one bucket by initializing the HashSet<T> with the first one's items in case there are a lot of items, it won't need to be resized.
+                            if (batchingBuckets.Count == 1)
                             {
-                                IList<TaskItem> itemsToAdd = bucket.Expander.ExpandIntoTaskItemsLeaveEscaped(targetReturns, ExpanderOptions.ExpandAll, targetReturnsLocation);
-
-                                foreach (TaskItem item in itemsToAdd)
+                                targetOutputItems = new HashSet<TaskItem>(batchingBuckets[0].Expander.ExpandIntoTaskItemsLeaveEscaped(targetReturns, ExpanderOptions.ExpandAll, targetReturnsLocation)).ToArray();
+                            }
+                            else
+                            {
+                                HashSet<TaskItem> addedItems = new HashSet<TaskItem>();
+                                foreach (ItemBucket bucket in batchingBuckets)
                                 {
-                                    if (!addedItems.Contains(item))
-                                    {
-                                        targetOutputItems.Add(item);
-                                        addedItems.Add(item);
-                                    }
+                                    IList<TaskItem> itemsToAdd = bucket.Expander.ExpandIntoTaskItemsLeaveEscaped(targetReturns, ExpanderOptions.ExpandAll, targetReturnsLocation);
+                                    addedItems.UnionWith(itemsToAdd);
                                 }
+
+                                targetOutputItems = addedItems.ToArray();
                             }
                         }
                     }
                 }
                 finally
                 {
                     // log the last target finished since we now have the target outputs.
-                    targetLoggingContext?.LogTargetBatchFinished(projectFullPath, targetSuccess, targetOutputItems?.Count > 0 ? targetOutputItems : null);
+                    targetLoggingContext?.LogTargetBatchFinished(projectFullPath, targetSuccess, targetOutputItems.Length > 0 ? targetOutputItems : null);
                 }
 
-                _targetResult = new TargetResult(targetOutputItems.ToArray(), aggregateResult, targetLoggingContext?.BuildEventContext);
+                _targetResult = new TargetResult(targetOutputItems, aggregateResult, targetLoggingContext?.BuildEventContext);
 
                 if (aggregateResult.ResultCode == WorkUnitResultCode.Failed && aggregateResult.ActionCode == WorkUnitActionCode.Stop)
                 {