Rate-per-second gauge meters derived from cumulative samples

Currently we capture several detailed metrics related to disk & network I/O from the hypervisor in cumulative form, specifically:

* disk.read.requests
* disk.write.requests
* disk.read.bytes
* disk.write.bytes
* network.incoming.bytes
* network.outgoing.bytes
* network.incoming.packets
* network.outgoing.packets

These values being cumulative are not suitable for alarming against with the static style of alarm threshold that ceilometer supports. Cumulative values are not best suited to charting either via the new metering dashboard as trends are less easy to spot visually.

However we could follow the pattern used to derived cpu_util from cpu samples via the rate_of_change transformer to produce derived meters based on the rate of bytes or packets flowing per second. Such gauge meters would be much more useful from an alarming and dashboarding perspective.