Driver retries TRUNCATE_ERROR (0x1003), but shouldn't

[gusev-p]

The truncate_statement class converts any error happened during truncate table execution to TRUNCATE_ERROR. The driver calls retry_policy.on_request_error for TruncateError which returns RETRY_NEXT_HOST, this doesn't make much sense. For example, when a table is dropped during truncate table the user will see unconfigured table <table name> -- the validation triggered on the second retry, instead of the original error.

Another problem is that TruncateError Python class ignores the error message sent by the server -- it always returns unhelpful 'Error during truncate'.