fix(api): Fix certain failed runs showing a stale recovery target

[SyntaxColoring]

Overview

Closes EXEC-503.

Test Plan and Hands on Testing

I think unit tests are sufficient here.

As far as I know, this bug no longer has any user-visible effect that we can test against. This is because the PAPI layer handles error recovery much more comprehensively now, so this part of EXEC-503's steps to reproduce can't happen in practice:

2. The Python API naively lets the exception propagate, not taking advantage of the error’s recoverability

Changelog

The bug was that I added code to do the right thing in StopAction, but I neglected to add the same code in FinishAction.

When a run ends:

• It sometimes, but not always, goes through an opentrons.protocol_engine StopAction, depending on the circumstances.
• It's always goes through an opentrons.protocol_engine FinishAction.

Review requests

None in particular.

Risk assessment

Low.

[sfoster1]

whoops lol