encoding/dot: new package DOT graph encoding

[kortschak]

Currently, the package handles our basic graph types without any problem, but there are features in DOT that we can use (the additional interfaces do this - though they are not yet tested), e.g. subgraphs.

When this has been iterated over, I will write a DOT parser to do the unmarshaling (this has some hairy bits that I'm still thinking about, but it's less of a concern - I just want something to make our packages a bit more friendly for visualisation).

@gonum/developers

[kortschak]

This is ready for some review.

Current issues are:

1. We don't check for valid DOT ID syntax.
2. We don't really properly handle subgraphs - the addition of the strict option is to deal with that (we just pass the problem on to GraphViz). I don't think this is a huge issue since subgraphs are unlikely to be a common use case, and we can (I think at this stage) legitimately just force all graphs to not be pseudo-graphs (note that I'm still not convinced that pseudo graph support is not a worthwhile goal in the long run) by adding strict. Maybe fixed.
3. The code could be a lot cleaner.

PTAL

[kortschak]

ping

[vladimir-ch]

global?

[kortschak]

Good catch. That should not be there.

For reference, all this comes from this and playing around.

[vladimir-ch]

I think it LGTM.

[kortschak]

PTAL for the documentation wording.

I guess we can clean up the code at a later stage. It's reasonably readable, but I do think it could be improved upon.

[vladimir-ch]

I like that this is stated explicitly.

[vladimir-ch]

Much better now. LGTM.