changed logic for how errors are handled on the initial query in the wrangler.
For scripts that dont use the sql editor, if an error occurs during the first read of the data the error is lost. Kylo will redirect back to the build step/canvas and there is no indication of what went on.
Also no matter what the user does (i.e. remove the dataset and add an entirely new one) the Next button will always be disabled preventing the user from doing anything
I think with it was changed to use the HttpBackendClient, thus removing the snackbar for errors in the interceptor. We probably need that back and then handle the one off use case for the sql transform that doesnt want that.
Note We should also test inline transform errors and see how that works with the wrangler after the change
To test use the provided
The file has an attribute with a "." in it, which causes errors with the wrangler.