Fix nasty bug in CodeWriter's TextReader implementation

`CodeWriter.Reader.Read(char[] buffer, int index, int count)` is implemented incorrectly when called after called after it has no more data to return. In that case, it returns -1 rather than 0. Because this method is called in a loop by `TextReader.ReadBlock(...)` while the result is greater than 0, -1 does end the loop. However, It causes -1 to be *added* to the final length returned by `ReadBlock`, reducing it by 1. This causes the `LargeText` returned to have a length that is 1 fewer than expected.

In addition, the `TextReader` implementaion failed to reduce `_remainingLength` as expected. This is now fixed as well.

Author: DustinCampbell

src/Compiler/Microsoft.CodeAnalysis.Razor.Compiler/src/Language/CodeGeneration/CodeWriter.cs

@@ -415,7 +415,7 @@ public override int Read(char[] buffer, int index, int count)
 
             if (_page is null)
             {
-                return -1;
+                return 0;
             }
 
             var destination = buffer.AsSpan(index, count);
@@ -461,6 +461,11 @@ public override int Read(char[] buffer, int index, int count)
                         charIndex = 0;
                     }
 
+                    if (source.IsEmpty)
+                    {
+                        continue;
+                    }
+
                     source.CopyTo(destination);
                     destination = destination[source.Length..];
 
@@ -497,6 +502,8 @@ public override int Read(char[] buffer, int index, int count)
                 _charIndex = -1;
             }
 
+            _remainingLength -= charsWritten;
+
             return charsWritten;
         }
 
@@ -551,6 +558,7 @@ public override string ReadToEnd()
             _page = null;
             _chunkIndex = -1;
             _charIndex = 1;
+            _remainingLength = 0;
 
             return result;
         }