Support extraction functions

[jonross]

For when jmespath isn't enough, the user should be able to supply a function name.
These can be defined in ~/.kubeql/functions.py.
In fact, internal extractors like job status, cpu requests etc should be defined this way, using an internal functions file.
Could get rid of custom table builders?