* for all Illumina pre-processing methods the order of steps has been changed to now 1) search for primers/trim and then 2) merged PE reads. This ensures that all primer/adapter sequences are trimmed from dataset, whereas previously AMPtk releases merged PE reads first - because of how usearch/vsearch merge PE reads, occasional primer/adapter sequences were slipping through the pre-processing steps. Also reads with multiple primer hits (i.e. two forward primers) are now discarded. While this results in the pre-processing steps being slightly slower in runtime, it increases the data quality downstream.
* read orientation is tested/fixed on the fly for `amptk illumina2` workflow (barcodes/primers in reads). Some datasets of 50/50 read orientation.
* added a check for "inverted OTUs" for all denoising/clustering steps. This was largely a check to validate that changing illumina pre-processing steps were working correctly (as unintended result was a small number of OTUs that were on the "crick" strand)
* Default mapping file now has a 'RevBarcodeSequence' column. Mapping files used for `amptk illumina2` enforce the paired barcode sequences in the mapping file. If barcode fasta files are given, then all combinations of barcodes (5' and 3') are saved.
* a few py27/36 bug fixes