Write job-specific subsampling config files

[victorlin]

Note

These changes are now part of #90 (efe1483...c4ecba6) – see #91 (comment)

Description of proposed changes

This prevents unrelated config changes from triggering a re-run of augur subsample.

Related issue(s)

• Improve input block when using augur subsample public#31

Checklist

• Checks pass
• Update changelog

[joverlee521]

Overall changes LGTM, small comment about invalid build errors.

[joverlee521]

non-blocking nit

Should we start standardizing the use of InvalidConfigError that is included in the vendored config.smk?

Suggested change

	print(dedent(f"""\
	ERROR: The following names in 'builds' are not valid:
	{indented_list(invalid_builds, " ")}
	Valid builds are:
	{indented_list(VALID_BUILDS, " ")}
	"""), file=sys.stderr)
	exit(1)
	raise InvalidConfigError(dedent(f"""\
	The following names in 'builds' are not valid:
	{indented_list(invalid_builds, " ")}
	Valid builds are:
	{indented_list(VALID_BUILDS, " ")}
	"""))

Changes output to

$ nextstrain build phylogenetic/ --config builds='["foo"]'
InvalidConfigError in file "/nextstrain/build/phylogenetic/rules/config.smk", line 30:
The following names in 'builds' are not valid:

    foo

Valid builds are:

    genome
    N450

  File "/nextstrain/build/phylogenetic/rules/config.smk", line 54, in <module>
  File "/nextstrain/build/phylogenetic/rules/config.smk", line 15, in main
  File "/nextstrain/build/phylogenetic/rules/config.smk", line 30, in validate_config

[victorlin]

+1 for standardizing, but I prefer the simpler error without stack trace at the bottom. How about catching InvalidConfigError from main()? I've done this in force-push.

[victorlin]

GitHub did something weird here. I wanted to swap the order of #90 and #91 since it makes more sense, so I rebased locally and force-pushed. It looks like GitHub detected this as #91 being "merged" into #90, so I'll just roll with it... it aligns more with nextstrain/WNV#116 anyways.