backend: indefinitely retry workers' talk to frontend
... again. I accidentally dropped the FrontendClient flag when moving
things around in b7f71e173b26083781562e44b11034a41c4fad52.
The weird effect of this bug was when we had a small FE downtime today.
The backend workers finished several builds for iucar/cran project - and
then failed to mark the jobs as finished on FE due to the FE downtime.
The background worker ended, the corresponding process doesn't exist
anymore -- but the builds still look like "running" on FE.
Since the project is still receiving many other build requests (at the
time of writing this patch), and all those need source builds first
(srpm builds are prioritized over normal RPM builds) those _already
finished_ builds are still around in "running" state (several hours
now). They are waiting till they bubble up in the queue again.
Fixes: #1413
Merges: #1412
Relates: #1414