TODO

Enhancement requests
--------------------
* For every description add the exact kernel counter involved w/ proc path and also
  add a proper comprehensible explanation. For example in 'ldavg-1' add explanation
  when a load is high and how it relates to the number of CPUs

* Normalize the y-scale
  Currently the maximum value on the y-scale is simply the maximum value of the graph.
  Add an --option (--normalize-max) that makes the maximum always the same so that
  different graphs are easily comparable. We'd need to define "similar" graphs and
  choose the maximum of the graph among all the similar graphs. For example:
  all % graphs will have the max at 100. If an sosreport is present, we could use
  certain values there to define the max. For example: /proc/meminfo will give
  us the maximum amount of RAM. Use that to set the maximum. Suggested by Luca
  Miccini.

* Allow --ascii graphs to have multiple datasets
  Currently we only allow one dataset in the --ascii output

* Add multiple --graph options that work just like --svg and create a single short
  pdf file with the listed graphs

* In custom graphs it should be possible to specify a custom time interval

* Implement a --range option that highlights the area between two times/dates

* Resolve devX-Y to proper device names via sosreport parsing