Capture corrupted binlog events at runtime

Apart from binlog media failures, corrupted events can be caused by malfunctioning
server itself.
There were cases when the server binlog was exposed as a target of
misplaced writes to. E.g by a faulty plugin.

This task aims to discover facts of corruption.
A legal binlog (e.g group-commit) writer will try reading out,
as pre-write-catch,
few common headers (19 byte each) of last
events binlogged before to submit or write to binlog own event(s).
At writing to binlog the writer would memorize offsets of the being written events.
At completion respective headers would be tried out reading back - Post-write-catch.

When any error is detected some actions could be done, this needs
elaboration.