spec: conclude the consensus `overview.md` documentation

[cason]

The Tendermint consensus algorithm overview contains somehow our interpretation of the Tendermint paper and the associated implementation of the algorithm. It should not, however, dive into our implementation details.

However, it has some missing part and some TODOs:

• spec: Define the requirements for the proposer selection algorithm #396
• feat(spec): discussion of consensus getValue() function #760
• spec: Formally define the valid(v) properties #510
• feat(spec): Consensus network requirements and gossip property #764
• feat(spec): Consensus timeouts and synchrony assumptions #765
• Stretch goal: feat(spec): Argue that Tendermint is a safe and live consensus algorithm #777

[cason]

As also pointed out by @josef-widder, this document that should be an "overview" is getting super long and complex.

We should consider splitting it into parts, once its "full content" is defined.

[cason]

Another point to consider is the absence in the pseudo-code and the description of the algorithm in general of a decision primitive, to deliver decide values to the context/application.

The decision is achieved by appending to the decision_p array the decided value, the array being indexed by heights. This behavior is described in this section, but the actual interface with the context/application is not discussed.

[cason]

I am closing this issue because the protocol overview has been read and reviewed by different people.

@josef-widder, feel free to open an issue if you still want to dismember it into smaller documents, with different levels of complexity (the real basic and details).