The reader did not properly consume empty lines after =encoding commands, which produced various incorrect parses depending on the content between there and the next command. Fixes #10537
