The baseDir is the most important directory, because it directly affects JPATH
resolution.

It was assumed as the pwd so far. However, it is handy in scripts to specify it
on the command line. This behaviour switches to specify only.

BREAKING: You must specify the baseDir on the command line. To mimic the old
behaviour, use `tk show .`

Adds CLI-completion support to tanka 🎉 

The main workflow commands (show, diff, apply) now support completing the environment.
This is done using a combination of spf13/cobra and posener/complete.

Supported shells are bash, zsh and fish. Install using `tk --install-completion`

So far, diffing was offloaded to kubectl. While this produces very nice results,
it is only possible for kubernetes version 1.13+.

Nevertheless, there are older cluster versions around, so these need to be
supported as well. subset-diff addresses those cases in the hopefully best way
possible.

To reduce field bloat, it only diffes those fields, that are present in the
local config. Kubernetes adds dynamic fields on the fly which we cannot know
about, so this is required.

Note: You WILL NOT see removed fields in the diff output. Upgrade your cluster
version to 1.13+ and use native diffing.

Adds a detailed design document outlining the problem tanka is trying to solve and the taken solution.

Adds a tool (tk tool imports) to get a list of all files, a file transitively imports.

Optionally allows to check whether a git commit changed one of those imports.

BREAKING: tk debug jpath is now tk tool jpath

Addresses several open points that made using subset-diff impossible:

- produced a `panic` in case a slice returned from kubernetes was bigger than the one in the config
- showed differences on the apiVersion. While those were indeed valid, it was too much bloat so they are removed for now
- parallelizes generation of the subset. This includes parallel invocations of kubectl, which give an additional performance boost 🎉