fix: Retry on NeptuneRetryableError. Call on_async_lag_callback

[michalsosn]

NeptuneAsyncLagThresholdExceeded is a NeptuneScaleError, so when it was after NeptuneScaleError in the except block, the more general on_error_callback was called to handle this exception and on_async_lag_callback was a dead code

NeptuneRetryableError currently has 3 subclasses (NeptuneConnectionLostError, NeptuneUnexpectedResponseError, NeptuneInternalServerError)
NeptuneConnectionLostError is retried for 60 seconds with an exponential backoff, the other two subtypes simply triggered on_error_callback, which by default kills the sender process. It was possible to provide a custom on_error_callback to make them actually retriable, but with the default config they were simply fatal. I added them to the backoff instruction, so they are all handled in the same way, ie. retried for 60s and only afterwards the decision is delegated to the on_error_callback

[kgodlewski]

Just a tiny nit with test name