Found a better way to update the index after get/drop in v6 repositories. I was able to close all the todos around that.

Only problem is there is a race where a modification that happens to a file soon after get/drop gets unexpectedly staged by the index update. I made this race's window as small as I reasonably can. Fully fixing it would involve improvements to the git update-index interface, or another way to update the index.

Only two todos remain in ?smudge that I want to fix in the remainder of this v6 sprint.

This work is supported by the NSF-funded DataLad project.