Fix the Herbie egraphs so make works

[pavpanchekha]

I misunderstood the file format, whoops!

[pavpanchekha]

@mwillsey can you review & merge this? Basically I had enodes point to eclasses instead of having enodes point to enodes. This fixes #44

[oflatt]

The enode-eclass stuff keeps biting us
I wonder if we should have a separate file format for proofs instead of trying to make this one support both. Though I still like the idea of having a shared representation

[mwillsey]

I like keeping node-node. It's just more flexible. You can emulate node-class on top of it, but you can't go the other way.

[oflatt]

Gus ran into the issue that egraphs with cycles are not representable, though. The work-around is to keep fresh symbols in the egraph where you created the cycles in the first place

But otherwise I agree

[mwillsey]

That doesn't strike me as an inherent problem of this format, you can write down node-node cycles just fine. Put perhaps other parts are assuming some kind of acyclicity.

[oflatt]

Oh I see, they just won't be grounded terms