show free resources

[JRMeyer]

if you want to send a job to a GPU, you need to deduce which gpus are free via:

pit show nodes

and then:

for node in nodes:
    pit show node:$node

This isn't very efficient, and something along the lines of pit shoe nodes:free would be better

[tilmankamp]

This is not as trivial as it sounds, as the scheduler could have queued-up jobs that are waiting for some resource to be released.
So there are at least two different reports required:

1. Showing nodes and GPUs that are not used at the moment (rather simple to implement, but requires also changes to the Snakepit service itself).
2. Showing nodes and GPUs that are not used by running jobs and that no job is waiting for (rather hard to implement and also requires changes to the Snakepit service).

Admittedly in practice report 1 is enough and would help already.