scaffolding and reference selection based on ANI

[dpark01]

This PR introduces ANI-based mechanisms for reference selection using skani. This:

• updates viral-assemble to 2.3.1.1
• adds a task select_references to tasks_assembly.wdl that selects a subset of reference genomes based on ANI similarity to a set of given contigs/MAGs. This also clusters similar reference genomes to each other based on ANI and chooses only the top reference per cluster.
• changes task scaffold's behavior in task_assembly.wdl to chose a reference genome based on ANI when provided multiple reference genomes.
• changes the behavior of scaffold_and_refine_multitaxa to:
  • select a subset of reference genomes to scaffold against using ANI similarity to the contigs instead of a kraken-based reference selection
  • call scaffolding and refine steps on a per-reference-cluster basis instead of per-reference-taxon, forcing a choice of the most appropriate taxon of highly-related taxa
  • no longer fall-back to reference based assembly, since de novo scaffolding should always work if an ANI hit was found

This updated version of scaffold_and_refine_multitaxa should be far more efficient at metagenomic reference selection, produce less noisy outputs (less secondary taxon hits), still allow for diverse coinfections of unrelated taxa, and should perform well with a very large reference genome database as input.

This PR also introduces an unrelated change to modify the workflow terra_tsv_to_table to accept and concatenate multiple input tsv files.

[dpark01]

After a few runs on production data, the general methodological changes here look good. Some parameter tuning probably remains to be optimized in order to increase the clustering/matching between various rhino C species, but that can be experimented with separately and introduced in a later PR.