Add performance remarks

Author: koryboc

src/BinaryToText/Base85.cs

@@ -14,6 +14,7 @@
     using Resources;
 
     /// <summary>Provides functionality for encoding data into the Base-85 (also called Ascii85) text representation and back.</summary>
+    /// <remarks><b>Performance:</b> Highly optimized. Base85 encodes data in independent 4-byte groups, forming no serial dependency chain between groups. This makes it highly amenable to SIMD vectorization (AVX2) and parallel processing across multiple cores, resulting in throughput that scales with available hardware.</remarks>
     public sealed class Base85 : BinaryToTextEncoding
     {
         private const int ChunkSize = 1024 * 1024;

src/BinaryToText/Base91.cs

@@ -8,7 +8,8 @@
     using Resources;
 
     /// <summary>Provides functionality for encoding data into the Base-91 (formerly written basE91) text representation and back.</summary>
-    /// <remarks>See more: <seealso href="http://base91.sourceforge.net/"/>.</remarks>
+    /// <remarks><b>Performance:</b> Poor by design. Base91 is inherently sequential — each output word depends on an accumulated bit-state carried over from the previous byte, forming a serial dependency chain. This makes parallel encoding and decoding impossible without a full pre-scan pass that replicates the entire computation, which negates any potential gain from parallelism.</remarks>
+    /// <seealso href="http://base91.sourceforge.net/"/>
     public sealed class Base91 : BinaryToTextEncoding
     {
         /// ReSharper disable CommentTypo
@@ -135,7 +136,7 @@ public override void DecodeStream(Stream inputStream, Stream outputStream, bool
                     db[1] = -1;
                 }
                 if (db[1] != -1)
-                    bso.WriteByte((byte)((db[2] | (db[1] << db[3])) & byte.MaxValue));
+                    bso.WriteByte((byte)((db[2] | db[1] << db[3]) & byte.MaxValue));
             }
             finally
             {