more scaffolding updates

[dpark01]

scaffold_and_refine_multitaxa:

• more resilience to empty outputs or partial recovery of multi-segment genomes
• update input reference genome data structure to tsv instead of json string
• auto-filter input reference genome list (optionally) based on kraken hits
• run polish only if scaffolding successful--if not successful, fallback to refbased assembly as hail mary attempt
• more top-level (terra table) outputs

classify_single:

• more top-level workflow outputs summarizing primary viral taxa found, and how much

containers:

• bump viral-classify 2.2.3.0 to 2.2.4.0

build:

• update cromwell/womtool version

[tomkinsc]

It may be nice to break out tax_name and percent_reference_covered for the "top" viral assembly into separate workflow outputs, for easier search and filtering on Terra (where "top" could be defined as the most complete assembly, or the most abundant taxon in terms of # of reads or # of matching distinct k-mers).

[dpark01]

Added into the TO DO comments at the bottom of the WDL. I think this will require a small bespoke tsv-parsing task for this purpose. It will also need to be reslient to the empty-output scenario (ie, there is no top assembly because none were attempted or were successful).

[tomkinsc]

Any reason we can't make this type Map[ String, Map[String,String] ], where the outer map String keys are the taxid or tax_name values? (for picking out values for a given taxon in downstream analyses)

[dpark01]

Mostly just because of how we construct it (see the scatter in the WDL above), and that WDL 1.0 lacks a lot of the basic methods for navigating Maps and converting back and forth with Arrays.