Bug fix for `sequence` method of FASTA Records

Previously if the user did not specify that the version of sequence they wanted came from the FASTA namespace (which is not always possible if writing generic code to work with any record reader type we offer), then a version of sequence which does type prediction would always be used, resulting in highly type unstable code, puzzling to some first time BioJulia users. This fixes that so as the sequence type can be specified, allowing type-stable efficient code generation.