Input can be a path prefix

[nh13]

The user may provide a path prefix (or directory) in which undetermined FASTQs live. The file names are assumed to be of the form: <*>_L00#_<R# or I#>_001.fastq.gz. The following can be inferred from this pattern: the lane number, read number, and if the FASTQ is a template or index read.

If no read structure is given either on the command line or Sample Sheet, then a new list of read structure is built, one per input FASTQ, where the kind (segment type) is inferred from the file name and all bases are used. E.g.

test_L001_I1.fastq.gz
test_L001_R1.fastq.gz
test_L001_R2.fastq.gz
test_L001_I2.fastq.gz

would yield the read structures +B +T +T +B. Reads will all molecular barcodes are supported (e.g. test_L001_U1.fastq.gz) and bases to skip (e.g. test_L001_S1.fastq.gz). The utility of the latter is questionable, but kept for completeness.

Improvement: due to the need for all sample barcode read segments to have a fixed length (for index hopping metrics collection and some validation), the first read is read from each FASTQ that contains a variable length sample barcode to determine the read length and fix the length of the variable length sample barcode. I don't believe that a read structure like +B +T +T +B previously worked!

[tfenne]

A few minor comments and suggestions, but good to merge once you're happy.