Output keywriter log, provisioning log and die messages to stderr

[oliora]

In our workflow we run rpi-sb-provisioner.sh manually and if anything went wrong it's inconvenient to gather logs from multiple places.

For die function it looks more logical to also output the die reason to stderr rather than to debug "channel".

[tdewey-rpi]

Thanks for the submission @oliora

Having considered the change and implications, I am not inclined to accept this PR as-is:

1. I only have the bandwidth to support rpi-sb-provisioner in the flow as described (automated hands-off with complete control).
2. The mechanism you're using, as it is not supported, is likely to be broken imminently by our upcoming 2.0 release, which will introduce multiple changes to the provisioning flow for easier debugging, faster flashing, and more features
3. This change could also have been implemented by testing $DEBUG, and then redirecting the messages to that location - which would have been something I could forward-port without worrying about expanding the support cone.

[oliora]

@tdewey-rpi fair enough but still would be useful if all logging is duplicated to a single place unless it's something that is not trivial to implement. Even if rpi-sb-provisioner.sh is not supposed to be called directly its output may be observed via journalctl to troubleshoot.

Please also remember that it's not possible to provision RPi5 automatically since it does not have a jumper for OTG boot so provisioning needs to be assisted with rebooting the device in OTG mode when needed.