Improve configuration serialization in `.headache-run`

[fbiville]

headache currently needs to track two pieces of data:

• the (Git) revision it last ran at (it assumes .headache-run was committed along with the source file changes)
• the configuration that was used then (via a configuration:/path/to/headache.json entry in the headache-run) that allows headache in a subsequent run to read the former configuration and header

The downside of this is that headache may not work with shallow clones as it requires access to enough of Git history to be able to read configuration at the revision during which headache last ran.

[fbiville]

One solution would be to serialize both the JSON configuration and the license header text in .headache-run so there is no need to dive into the Git history to get those elements next time headache runs. That way, headache would still work fine with shallow clones.

[fbiville]

And that would probably fix https://github.com/fbiville/headache/issues/55 as a nice bonus effect!