Track client reset reason in table that detects client reset cycles

[sync-by-unito]

Right now we track when an automatic client reset begins, and if we start another automatic client reset before ending the first, we raise an error that looks like A fatal error occured during client reset: 'A previous '<reset mode>' mode reset from <date> did not succeed, giving up on '<reset mode>' mode to prevent a cycle'. We should also track the reason the first client reset began so that we can see in the second aborted client reset what error caused us to begin the cycle in the first place.

[sync-by-unito]

➤ michael-wb commented:

This fix was originally bundled with the fix for RCORE-2063, but has been separated out into a different PR. Once this has been landed, RCORE-2063 can be completed.