similar to --input-dir option, it is worthwhile to allow users to specify an input directory with the executables. while this is not a full blown transformation catalog replacement, it will lower barrier for new users to get started

in terms of what arch and os are associated with the executables, the planner will make them match the compute site to which the workflow is being planned against

If there are one or more compute sites, then the planner should throw an error if the architecture of the sites dont match