Fix the upgrade failure from any Mender client version "<=3.2.1" to ">=3.1.0"

This is a multi factor problem which requires several fixes to work together. It happens because the client does not store the artifact name when it is first provisioned, it only reads it from /etc/mender/artifact_info. When a rootfs-image deployment happens, once the client starts on the new partition, it has no record of this artifact_name anymore, and can only send the new one, which is in the new /etc/mender/artifact_info. However, the deployments/next endpoint expects the client to send the same name as in the previous call, namely the old one.

Rather than trying to patch this with various fixes, we are going to change the design, so that the client will call a new endpoint to get Update Control Maps, instead of re-using deployments/next which is not really meant for this. This new endpoint will not need the artifact_name at all.