Handle Errors in Change Pipeline

[jeffbcross]

If something goes wrong at any stage of the change pipeline, there's nothing that gives the developer an opportunity to handle errors.

For example:

• What if the protocol rejects a model update?

SyncResource should provide a flexible means of reacting to things going wrong.

[jeffbcross]

A couple of ideas came out of an IRL conversation with @ritch about this.

• Maybe a middleware in the change pipeline could implement a special error-handling signature, and would automatically get called if any part of the change pipeline called next("some error");
• OmniBinder could provide a library of out-of-the-box error-handling middlewares, such as one that reverts a model if the change failed to be applied to a protocol.