read_merger.pl to support named pipes as inputs

[clintval]

Would you be willing to support named pipes when classifying with Kraken?

I store paired-end reads in BAM and, instead of storing the intermediate FASTQ on disk, stream the paired sequences into Kraken with named pipes. This offers good speedups as I am not writing to the filesystem when using BAM as a source file. Thanks!

Example:

❯ mkfifo /tmp/fastq{1,2}
❯ picard SamToFastq \
    INPUT=input.bam \
    FASTQ=/tmp/fastq1 \
    SECOND_END_FASTQ=/tmp/fastq2 \
  | kraken \
    --fastq-input \
    --paired \
    --db "${DBNAME}" \
    /tmp/fastq1 \
    /tmp/fastq2 \
    > output.txt

[clintval]

Actually. I just realized this is already in PR #89. I will wait for that!