Refactor EVM integration

[enlight]

Right now to get an instance of the EVM we do NewLoomVm()->NewLoomEvm()->NewEvm()->NewEnv()->go-ethereum/core/vm.NewEVM() and loomchain.State is passed through these layers all the way to NewEvm() making it hard to reason about what kind of state changes can affect EVM calls.

The EVM integration code should not have direct access to loomchain.State, and it shouldn't be directly accessing feature flags nor on-chain config settings via that state, any configuration should be explicitly specified when creating the top level EVM wrapper (i.e. NewLoomVm()).