Update petgraph and use StableGraph

[Jengamon]

Hello.
I'm here, getting ready to use daggy in a project of mine, when I check Dag's documentation, and find out that removals may change indices. I looked around (docs.rs), and found that because daggy used petgraph 0.2.2, it missed out on petgraph 0.4.3's StableGraph.

I would like to propose updating, and building Dag off of StableGraph, in order to stablize indices.

[mitchmindtree]

@Bobhostern Something along these lines sounds great 👍 I think I'd prefer if we maintained both options somehow: the current slightly more performant option as well as a version based on StableGraph.

I've been thinking about this for a while and I'm still unsure what the best abstraction for this is. Personally I think I'd like to see petgraph's Graph be generic over the node/edge storage so that it did not require an entirely separate, almost identical API implementation just for the StableGraph. However I understand that this is tricky to do cleanly without something like Higher-Kinded Types. As a result, it's probably best for us to do something similar in daggy where we implement a second, StableDag type.

I'm slightly tempted to just switch the Dag to use StableGraph internally, however it is hard to tell how much this would impact downstream crates like dsp-chain and conrod.

[bluss]

I gave that a big attempt before picking the current way. There's a lot more traits that G & SG have in common now, though, than before.

[vilunov]

Any progress on updating petgraph dependency? Recent std changes made petgraph broken, which was fixed in 0.4.5.

[azriel91]

#16 doesn't address the changing indices, it just allows daggy to compile.

[mitchmindtree]

daggy 0.5.0 was just published which now depends on petgraph 0.4.5, fixing the issue on nightly. I'm going to leave this issue open though as I still think it would still be nice to add a StableDag type.

[davenza]

I just opened a pull request with an early implementation of the StableDag structure in #21.

[mitchmindtree]

Closed via #21.