Refactoring integration tests

Current integration test setup is very brittle and is not assuming full working blockchain. On top of this there are too many assertions unit testing what should already have been tested. It means that every time we add some new feature or change some feature, we can either break a lot of tests or in the opposite direction have unexpected consequences and possibly not see some changes happening. I am proposing to:

- [ ] Setup proper chain environment that mimics real chain with proper block initialization and finalization sequences
- [ ] Have proper state transition functions taking into account all of the internal interactions
- [ ] Have a set of pre-defined transactions ready that will result in expected scenarios
- [ ] Use all of the above to re-construct a set of lean integration tests without side effects and hacks
- [ ] Improve test coverage while not testing unnecessary interactions which are covered by unit tests