EVM Proxy-server: split the mockup kernels in two phases

In the design made by @lthms:

The execution is in two phases:

1. Collect the block's propositions and fill the delayed inbox
2. Execute the pending blocks queue

We want to have this worfklow in the EVM mockup kernel so others can potentially plug their real-world code.

I believe for our needs we can say that:

1. A block's proposition is a list of transactions
2. We have one block proposition per tezos level