calling dc_backup_provider_new twice lacks last error set

[r10s]

if the user cancels the creation of backup_provider_new(), UI is supposed to call stop_ongoing_process.

this canceling, however, seems to take some tens of seconds sometimes - if the user attempts to create another backup provider during this time, this fails with the log line There is already another ongoing process running..

however, the logged line is not reported as an error, get_last_error() just reports nothing that could be shown to the user - i assume, we need to call error! instead of ensure! - error sets last error automatically afaik. the issue is probably also true for other ongoing processes, however, did not really appear there.

finally, of course, in addition, also a faster abortion of backup_provider_new() would be nice, but this issue is only about error handling (there could also be other errors)

related to #4007