Improve input block when using augur subsample

[victorlin]

Progress

• subsample: Improve relative filepath search augur#1897
• config: Set AUGUR_SEARCH_PATHS shared#68
• Write a specific config section shared#64
• Update subsample inputs docs.nextstrain.org#266
• Update subsample inputs measles#90
• Update subsample inputs WNV#116

Background

When a filepath is defined in Snakemake's input block, The mtime rerun trigger allows Snakemake to intelligently¹ rerun the workflow upon modifications to the file.

¹ The mtime trigger is also conditional on the file contents being changed (src). Note this only applies to small files (i.e. under --max-checksum-file-size, default 1MB).

Description

Currently, input blocks for subsampling rules in repos using augur subsample are problematic in two ways:

1. Every config change, even to unrelated config sections, triggers a re-run of augur subsample.
   • This is because the entire config dump is specified as an input.
2. Changes to files referenced within the subsampling config do not trigger re-runs of augur subsample.
   • This is because the referenced files are not specified as inputs.

Possible solutions

1. Dump a subset of config (the section used for augur subsample) per invocation and specify it as input.
2. Add a helper function to return a list of referenced files in augur subsample config.
   • This should consider filepath resolution for compatibility with nextstrain run.