feat(api): Enable error recovery without the feature flag

[SyntaxColoring]

Overview

Closes EXEC-547.

Test Plan

• Flex protocols still enter recovery mode when there's a tip pickup or overpressure error, and still fail for other errors.
• opentrons_simulate, local in-app analysis, and robot-side analysis still work as expected when the protocol has an error.
• OT-2 protocols still fail, without entering recovery mode, when there's an error.

Changelog

• Stop gating error recovery behind the enableErrorRecoveryExperiments feature flag.

  The feature flag doesn't do anything anymore. I'm leaving it in-place in case we want to use it for the ongoing work in liquid level detection.

• For things other than real robot-server protocol runs (including analysis, opentrons_execute, maintenance runs, etc.), explicitly choose a policy of "never recover."

  We don't ever want to enter recovery mode for these things, because nothing could bring it out of recovery mode, and it would hang forever.

• Explicitly disable error recovery on the OT-2. None of the designs for the upcoming release make sense on an OT-2, because the OT-2 lacks the hardware feedback to detect the errors in the first place.

Review requests

• I looked around for a frontend feature flag but I couldn't find one. Have we already removed it?
• Anyone have thoughts on what error recovery policy is appropriate for stateless commands (the "default orchestrator", in this code)?

Risk assessment

Medium.

[sfoster1]

Code looks good to me, haven't tested though. I think never_recover is appropriate for stateleess command execution.