disk utilisation: use proc fs and extend metrics

[cha87de]

• to be extendable for containers, rather use the proc fs than reading the values from libvirt daemon.
• missing metrics: wait times, queue sizes, cache misses, errors, etc.

[cha87de]

moving to proc fs as source seems to be not possible, since block device may be remote. The IO collector gives identical information and should be used as an alternative when the disk i/o should be read from the proc fs.

remaining open tasks: add host metrics to detect bottlenecks (max bandwidth, etc.)