Double (invalid) slashing tests

[protolambda]

Tracking isssue, we should test double slashings. Lighthouse had a bug here, caused by over-optimization that was unaccounted for in the spec: sigp/lighthouse#1065

The tests should cover:

• Two proposer slashings in the same block, for the same proposer, making the block invalid
• Two attester slashings with overlap in attesters in the same block, similar to above.
  • Valid case: different sets of attesters with overlap, repeated in two attester slashings, but overlap should not be slashed twice.
  • Valid case: one attester twice, no other participants

[djrtwo]

I'd like to leave this issue open even after the close of #1781

I think it worth considering upping MAX_ATTESTER_SLASHINGS to 2 in v0.12. The main reason to keep it low is due to size considerations (max ~33kb per AttesterSlashing when 4M validators), but in lower ranges of validators (say ~300k validators) this is more like ~3kb.

The other added benefit is that it takes half the amount of time to fully slash an attacker when a serious attack is found

cc: @vbuterin for thoughts on upping the value of this constant

[protolambda]

@djrtwo Can we close this issue now, with the tests in #1794 ?

[djrtwo]

yep!