Small fixes for problems spotted during flood testing.

Most important change is the one regarding future witness schedule object. Instead of assuming it will throw on first access and handling it when it happens to use "lazy init", now the object is created on activation of HF26 as a copy of current schedule. Then we can assume it is always present. Other changes are fixes for copy-paste bugs, some logging changes, and also the expiration time for fast-confirm transactions is no longer full hour, because if such transaction reaches other node before it processed the confirmed block, it would be considered invalid as expiring too far into the future.