Avoid retries for overly large lock requests

[sandorw]

Internal reference: 77636

We should avoid retrying failed lock requests where the request fails because of a 413 - retries aren't going to help, and may actively overwhelm Timelock.

[sandorw]

As per square/okhttp#1001, we don't even realize that it's a 413 and get a socket exception instead. Unfortunately, our retry logic on top of that is faulty (okhttp is configured to not retry at all at its layer) and means that we infinitely retry the request.

[sandorw]

As part of addressing this, another scenario comes up in PDS-77976

Clients retry timed out lock requests (especially on the legacy lock service where requests may block forever), but the lock service may respond to one of the timed out requests.