extend use of skewed clock to Transactional State facet implementation.

[sebastianburckhardt]

No description provided.

[jason-bragg]

Akk, sorry I didn't see this :/
Updated clock skew tests with suggested change which mostly duplicates this work, with some notable differences.

• Added Merge function to casual clock, because the logic of merge and mergeWithUtcNow are not the same.
• didn't use clock for rest of timing, as i was only concerned with transaction timestamps, and the skewed clock does not (currently) return time with a consistent skew, that is, each call to it gives random time within now+- skew.

[sebastianburckhardt]

• Having a separate function for Merge is fine. I was just lazy and the difference is not really relevant.

• I would prefer to use the injected skewed clock for ALL time management. It should all work - timeouts may be not be handled with great accuracy when someone is messing with the system clock at that moment, but it all must still work correctly and resolve without problems.