Use different default pixel sampling strategy for PNG

Author: JimBobSquarePants

src/ImageSharp/Formats/Png/PngEncoder.cs

@@ -3,6 +3,7 @@
 #nullable disable
 
 using SixLabors.ImageSharp.Advanced;
+using SixLabors.ImageSharp.Processing.Processors.Quantization;
 
 namespace SixLabors.ImageSharp.Formats.Png;
 
@@ -11,6 +12,16 @@ namespace SixLabors.ImageSharp.Formats.Png;
 /// </summary>
 public class PngEncoder : QuantizingImageEncoder
 {
+    /// <summary>
+    /// Initializes a new instance of the <see cref="PngEncoder"/> class.
+    /// </summary>
+    public PngEncoder()
+
+        // Hack. TODO: Investigate means to fix/optimize the Wu quantizer.
+        // The Wu quantizer does not handle the default sampling strategy well for some larger images.
+        // It's expensive and the results are not better than the extensive strategy.
+        => this.PixelSamplingStrategy = new ExtensivePixelSamplingStrategy();
+
     /// <summary>
     /// Gets the number of bits per sample or per palette index (not per pixel).
     /// Not all values are allowed for all <see cref="ColorType" /> values.

src/ImageSharp/Memory/Buffer2D{T}.cs

@@ -9,9 +9,6 @@ namespace SixLabors.ImageSharp.Memory;
 /// Represents a buffer of value type objects
 /// interpreted as a 2D region of <see cref="Width"/> x <see cref="Height"/> elements.
 /// </summary>
-/// <remarks>
-/// Before RC1, this class might be target of API changes, use it on your own risk!
-/// </remarks>
 /// <typeparam name="T">The value type.</typeparam>
 public sealed class Buffer2D<T> : IDisposable
     where T : struct