Handle fraud proofs and slash validators

[cwgoes]

Correctly handle fraud proofs from Tendermint and slash validators according to slashing conditions defined in the Cosmos specification (which will need to be unified / clarified in some cases).

Include tests somehow - CC @melekes recommendations for the best way to simulate fraud from the Tendermint side?

[melekes]

https://github.com/tendermint/abci/blob/master/specification.rst#beginblock

You can just construct abci.Evidence{height: 1, pubkey: pubKey.Address()}
https://github.com/tendermint/tendermint/blob/c28784de5e77095e89ed6480f655394a26497b1c/state/execution.go#L200-L203

I think we also want to include evidence itself because right now it's just pubkey and height. It's WIP pretty much

[melekes]

Refs tendermint/tendermint#1012

[cwgoes]

CC @rigelrozanski is this in the scope of #713?

[rigelrozanski]

The actual slashing will be performed by staking. however the fraud discovery doesn't need to explicitly a part of staking, a slashing keeper can be exposed by the stake module

[cwgoes]

Discussion at SDK design meeting:

• Simple double-signature evidence will be included in BeginBlock, Tendermint provides consensus on this evidence, so the SDK can modify application state
• Some kinds of fraud proofs may be transactions
• Complex evidence not implemented yet, may not be necessary at launch
• Need list of fraud proofs to be handled automatically at launch

[cwgoes]

Currently planning to implement just two slashing conditions at launch: double-signing and liveness (both of which will also automatically unbond).

[cwgoes]

Closing in favor of slashing v2 - #1117.