Allow asv run to consume list of hashes from a file

[eteq]

This PR enables the listing of hashes asv should run on by providing a file. Currently the format is intentionally limited to "one hash per line". The user says they want to use said file by giving "HASHFILE:xxx.yyy" as the range parameter to asv run. While not strictly a "range", this seemed ok because "ANY", "EXISTING", etc were already options, so this seems not too much farther from "range" than those.

The use case for this is wanting to compare a specific hand-picked set of commits to so how performance on a particular feature evolves with major known (also probably #561, if I understand the suggestion there correctly). To try to also address the use case #562 is going for, I added "stdin" or "-" as options, in which case stdin is treated as the "hash file", although if that's too hacky I could drop that commit.

I know I could do one of the tricks suggested in #562 instead of this, but those are a lot harder to grok, and also seem to preclude the use of some of the nice multi-commit features of asv run like --parallel or --interleave-processes.

Happy to add docs/etc if need be - just "testing the waters" for now, as I found it quite frustrating to not have an easy way to just run on a list of known hashes.

Closes #562

[pv]

Looks mainly ok.

Needs docs at least in the option help= text, and you can add it to test_workflow.py:test_run_spec.

Note also the CI currently has sporadic failures in some tests due to some conda issues unrelated to this.

[eteq]

Thanks @pv - I've pushed up a commit that I think addresses your inline review comments and adds to the test you suggested as well as some docs describing this feature in the range argument as well as the "using.rst" file.

[pv]

thanks, merged