The Issue struct is a cache of files on disk. There is never an Issue
without a directory to live in.
This commit adds a field to Issue to track what that directory is, so
that we can update those filew when we change the Issue, and commit the
changes to git.
This mostly provides an abstraction for "ephemeral worktrees", which
is a branch checked out in a worktree, to be read and maybe modified,
and the worktree is deleted/pruned when we're done with it.
There are also some helper functions for doing git things, the most
important one creates an orphaned branch.
The intent is to keep all the issues in a git branch. When we want to
do anything with issues (list them, add new issues, modify an issue,
etc) we check the issues branch out into an ephemeral worktree, modify
the branch, and delete the worktree.
