Introduce primary contigs

[pontushojer]

Contigs are now defined at three levels:

• all = every contig in reference
• primary = every contig in reference that should go through certain post-processing steps (see below). Is a subset of all
• phased = every contig in reference that is diploid i.e. can be phased. Is a subset of primary.

Post-processing steps run by primary contigs but not all:

• find_clusterdups + get_barcode_merges
• concat_molecule_stats + get_barcodes_to_filter
• call_variants

Reasons to add primary contigs include:

• You might not want certain contigs to be included in certain steps. For example in the 10x Genomics reference a decoy contig (e.g. hs37d5) which you would not want to influence barcode merging and filtering. Some contigs also have abnormally high coverage e.g. chrM. Others might just not be that interesting for your current application.
• get_barcode_merges and get_barcodes_to_filter are both bottlenecks in the pipeline where every chunk needs to be processed through to proceed. Thus the fewer contigs (and in extension chunks) there are the faster processing will take.

[FrickTobias]

Looks really nice. Since it was a really good description I would also copy paste part of your description in the PR into the permanent documentation docs/develop.rst.

Contigs are now defined at three levels:

• all = every contig in reference
• primary = every contig in reference that should go through certain post-processing steps (see below). Is a subset of all
• phased = every contig in reference that is diploid i.e. can be phased. Is a subset of primary.

Post-processing steps run by primary contigs but not all:

• find_clusterdups + get_barcode_merges
• concat_molecule_stats + get_barcodes_to_filter
• call_variants

[pontushojer]

@FrickTobias Thanks for the input! I have added a new section in develop.rst relating to the pipeline which contains the info.