removePrimers misses valid primer matches in certain circumstances

[gmagoon]

Currently, using removePrimers with orient=TRUE appears to miss valid primer matches in certain circumstances. Specifically, it seems that in the current removePrimers logic, if there is a fwd-only primer match on the original (non-reverse-complement) sequence, a full (fwd+rev) match on the reverse complement will not be considered. This can lead to unexpected behavior, e.g. "loosening" of matching thresholds can result in fewer passing read sequences. I believe #956 should fix this.