Acceptance criteria

Make it possible to:

1. Programatically and manually identify an activation from the .flox environment path
2. Manually delete all activation state without deleting the remote environment cache
3. Garbage collect activation state when a watchdog has died and the environment is no longer used

Why

Debugging bad activation state is hard because (corresponding to numbers above):

1. You can't inspect the relevant activations.json unless you can get $_FLOX_ACTIVATION_STATE_DIR from a working activation.
2. The crude option of deleting ~/.cache/flox/run to reset activation state also delete state for remote environments.

And:

3. We can't currently garbage collect the activations registry because we don't have an index of all activations or any mapping from the environment registry (create a follow-up ticket if not solved in this ticket)