Use one-shot UTF-8 encoding when possible.

This avoids allocating a EncoderNLS object and making multiple passes over the source data when the entire string could be converted at once. Additionally, UTF8Encoding.GetBytes is more likely to be highly optimized in the framework.

src/MySqlConnector/Protocol/Serialization/ByteBufferWriter.cs

@@ -99,11 +99,22 @@ public void Write(ReadOnlySpan<byte> span)
 		m_output = m_output[span.Length..];
 	}
 
-	public void Write(string value) => Write(value.AsSpan(), flush: true);
+	public void Write(string value) => Write(value.AsSpan());
 
 	public void WriteAscii(string value) => WriteAscii(value.AsSpan());
 
-	public void Write(string value, int offset, int length) => Write(value.AsSpan(offset, length), flush: true);
+	public void Write(string value, int offset, int length) => Write(value.AsSpan(offset, length));
+
+	public void Write(ReadOnlySpan<char> chars)
+	{
+		if (m_output.Length < chars.Length * 3)
+		{
+			var neededBytes = Encoding.UTF8.GetByteCount(chars);
+			if (m_output.Length < neededBytes)
+				Reallocate(neededBytes);
+		}
+		m_output = m_output[Encoding.UTF8.GetBytes(chars, m_output.Span)..];
+	}
 
 	public void Write(ReadOnlySpan<char> chars, bool flush)
 	{
@@ -273,7 +284,7 @@ public static void WriteLengthEncodedString(this ByteBufferWriter writer, ReadOn
 	{
 		var byteCount = Encoding.UTF8.GetByteCount(value);
 		writer.WriteLengthEncodedInteger((ulong) byteCount);
-		writer.Write(value, flush: true);
+		writer.Write(value);
 	}
 
 	public static void WriteLengthEncodedAsciiString(this ByteBufferWriter writer, string value)