Duplicate sample name case should be handled

[ressy]

The sequencers do allow a sample sheet to specify the same name for multiple samples, and currently this results in all but the first sample of a given name being "lost" during processing. How should this be handled? Based on Illumina's approach it should probably be supported, weird as it is. Maybe we should use sample number as the basis for tracking samples instead.

[ressy]

Still needs fixing, and using sample numbers would also sidestep the problem that the sequencer's onboard software and bcl2fastq have different naming conventions. Note that that Sample_ID in the sample sheet is not the same thing as sample number, and the output fastq.gz files appear to always be numbered sequentially.