Review op engine control algorithm about coweb

[vinomaster]

When complete, Readme file should have References section updated.

[parente]

Started this work in a branch. We were attempting to solve false-tie problem in transform functions. What we had might work in all cases, but it's not proven. For example, did not work in my quick experiments to hack in the op caching required for issue #12.

The COT algorithm we're using can side-step the false-tie cases though if we establish a total event order at the server and use it in the op engine to decide the order of transforms.

Implementation-wise the following changes are needed:

1. Server has to echo messages from site A back to site A in order. Right now, echoed messages are dropped.
2. Server has to assign an increasing integer to all /session/sync messages it publishes to a session establishing a total order of events.
3. The op engine has to sort operations in a context difference according to the order established by the server, not context order.
4. The insert/insert transform function should simplify back to position comparison with tie breaking based on site ID.

We should really look at cleaning up the protocol to make inserting these ids simpler, and only on user-to-user sync messages. Opening a separate issue for that.

[parente]

Blocked by #59 for now.

[parente]

Finished issue #59 so this is unblocked again. Continuing with this next.

[parente]

Updated JS, Java, and Python merged from a local branch (ot-41) into master. Completed work in 14a4a88 which includes an updated, interim coweb-0.5 JS build.

Net result of these fixes: the op engine implementation now breaks the CP2 pre-condition for operational transformation. This allows us to use the simple, proven, character-wise transformation functions that break ties based on site identifiers. This also opens the door for operation caching needed for issue #12.