When using perbase base-depth with --keep-zeros I sometimes get missing and duplicate loci in the output, which I think is the same as issue #71.

I believe this small change to the process_regions function in base_depth.rs fixes the issue which occurs when there are no-depth loci skipped in the result:Vector<PileupPosition>.

The previous code created new_result from result assuming the following situation:

                                                      |End
    -------------xxxxxxxxxxxxxxxxx---------------------
    |
Start

where the x's are where the result vector recorded non-zero depth positions.

The prior code first fills in the "left" of x's, then extends new_result with the x's from result, then
fills in the "right" of the x's.

But the code produces incorrect output when the situation instead looks like:

                                                      |End
    -------------xxxxx-xxx-xx-xxxx---------------------
    |
Start

I've added two new tests cases to check_depths by creating two new rstest #[fixture] called non_mate_aware_keep_zeros_positions and mate_aware_keep_zeros_positions. I'm not sure if this was the best approach and I'm definitely open to suggestions.

Besides these two test cases I've also tested on a small .bam file wget https://ftp.1000genomes.ebi.ac.uk/vol1/ftp/phase3/data/HG00096/alignment/HG00096.chrom11.ILLUMINA.bwa.GBR.low_coverage.20120522.bam with the following command which was producing dups/skips before the code changes but now lists each locus once.

cargo build --release

echo "11    4702317 4704317" > test.bed #tab sep

./target/release/perbase base-depth \
    --threads 4 \ 
    --keep-zeros \
    --bed-file test.bed \
    --output counts.tsv \
    HG00096.chrom11.ILLUMINA.bwa.GBR.low_coverage.20120522.bam