Test vectors

[Mischala]

Adding some test vectors may be helpful to ensure internal consistency.

• few vectors for expanding Secrets into Keypairs, and comparing a known result
• few vectors for signing a payload and comparing to a known result
• few vectors for verifying a result we know should correctly verify

[burdges]

We'll need them eventually of course. I'd accept PRs for them now. And we've commented out code that gives a starting point.

I've only just made a bunch of breaking changes, so test vectors would've made no sense before now. We could either add some now, as this might be the last batch of breaking changes, or else wait until we get a bit more feedback.

Ideally, we might do #26 before adding test vectors so as to suggest breaking optimizations. I halfway implemented #26 once before though so I think the only breaking optimizations is already done in https://github.com/w3f/schnorrkel/blob/master/src/vrf.rs#L736

[burdges]

In any case, I think benchmarks are a much higher priority. We'll always support merlin, and should've test vectors, but we might recommend alternative hashing with Shake128 or Blake2x instead of merlin. See #33

[burdges]

We should be ready to settle on test vectors around early August.

[burdges]

We've settled everything so this can proceed.

[burdges]

I'll close in favor of #43