Generator based testing framework

[danielSanchezQ]

Having generators that can deterministically generate can be used to do more complex unit tests.

When we have the ability to generate some of the chain types we would be able to generate more interesting test. For example, we could have a BFT blockchain generator. Conceptually this blockchain could have some specific block configuration and emulate a complex state which we can use to validates.
Another example of a test vector in where we can use this would be a generated blockchain that could be used to test recovery from storage in isolation. We could generate a simulated blockchain that would look like an original one generated by a node running for 10 days.

Ideally, the framework would:

• Support all the needed types from chain-libs.
• Be flexible and configurable enough to bend the generated values as needed.

Initial tasks:

• Support chain-core types
• Support chain-time types
• Support chain-addr types
• Support chain-crypto types
• Support chain-impl-mockchain types

[mzabaluev]

Consider bolero, an advanced fuzz and property testing framework; it could save us some cycles to not develop similar things from scratch, and it has a number of pluggable adaptive/heuristical fuzzers.