fixes to tbl_transfer_prealigned oob_clip

[dpark01]

Fixes and updates to tbl_transfer_prealigned --oob_clip behavior. Specifically:

1. CDS features are handled a bit specially: the 5' end is trimmed in multiples of 3 to retain original reading frame. 3' ends are still trimmed as-is.
2. oob_clip behavior was buggy for negative strand features: if the right end of the genome caused the 5' of a negative strand feature to be clipped, it was previously getting recorded as <1 which caused it to become positive stranded and encode the wrong half of the genome. This is now trimming to the right end of the genome instead (except with CDS features which get rounded off to the nearest multiple of 3).