Fix outer barcode matching: consensus from multiple reads, fuzzy matching, samplesheet-authoritative metrics

Three fixes for outer barcode (barcode_1/barcode_2) matching in splitcode_demux_fastqs:

1. Consensus barcode from multiple reads: Instead of trusting a single read
   header to determine the FASTQ outer barcodes, read the first N reads
   (default 10) and form a position-wise majority vote. This avoids failures
   when the first read has a mismatched index sequence.

2. Fuzzy matching with configurable mismatch tolerance: DRAGEN demux tolerates
   index mismatches, so FASTQ reads may carry barcodes that differ by 1-2 bases
   from the samplesheet. The new barcode_matches_fuzzy() function counts non-N
   mismatches and accepts matches within a threshold (default: 1). When multiple
   samplesheet entries match, the one with fewest mismatches is preferred.

3. Samplesheet-authoritative barcodes in output: Picard-style metrics and
   platform_unit values now use the samplesheet barcode values (the ground
   truth) instead of FASTQ-observed values that may contain Ns or mismatches.

New CLI parameters: --max_barcode_mismatches, --num_reads_for_barcode

Author: dpark01