Currently BlockFetcher performs bunch of validations on received responses but then it doesn't blacklist peer even if the response is clearly invalid.

The goal of this PR is to:

• refine validations performed in block fetcher
• restructure validation errors into proper error hierarchy (currently they are just strings)
• restore blacklisting, but make it depend on error returned (e.g. it's important to check whether received headers are forming chain with ones already enqueued but it should not result with blacklist, on the other hand if received headers are not forming chain themselves - it's clearly wrong response and it's worth blacklisting peer, who provided it)
• restore and extend test cases for blacklisting behaviour in Regular Sync