op-node: cleanup error-handling in derivation

[protolambda]

During the refactor of the derivation/driver code (See #10599) @ajsutton identified that the error handling is difficult to follow / getting dropped silently in some cases.

Separate from the initial refactor work we can change it to more clearly branch/return, and handle all error cases (even if the errors are not critical, at least they can be logged).

See:

• errors returned by TryBackupUnsafeReorg
• errors returned by TryUpdateEngine
• the return-if-not-EOF (nil or non-EOF error) cases