Fix empty_block_validation_regression test about casper-node

[piotr-dziubecki]

@Fraser999 to ping @afck about it

[piotr-dziubecki]

@afck : Well, it's a regression test. So it's fine to delete it as long as nobody reintroduces the bug. 😄
In this case I actually think it's fine to remove it. In fact, I'd remove the optimization from the code that allows skipping validation for blocks that are both empty and contain accusations. If all blocks are always getting validated without exception there's no danger of bringing back this bug.
So maybe just remove ConsensusValueT::needs_validation, and assume it's always true.

[Fraser999]

As discussed in standup today, we'll instead implement a fix for the test and retain the original optimisation in the code.