Factor out `seash` bits into `experimentlib`

[aaaaalbert]

Most of the functionality that seash provides for interactive use is useful for any other type of experiment manager, but difficult to use programmatically as-is because of how the current seash implementation ties in state and considerations of efficiency/user-friendliness (like keys, targets, parallelization, and success/failure tallying) into the functions that actually interact with remote nodemanagers.

(As an illustrative example, consider seash's and the current experimentlib.r2py's file upload functions.)

Instead of developing diverging code bases, let's factor out the valuable pieces from seash into SeattleTestbed/seattlelib_v2's experimentlib.r2py. This will require some thinking about disentangling functionality.

[aaaaalbert]

...and don't forget to look in the Clearinghouse codebase too! (The CH also splits, joins, sets users/owners, etc.)