Improves situation on LiteCore #437

[tleyden]

After this Sync Gateway change, the situation for https://github.com/couchbase/couchbase-lite-core/issues/437 incrementally improved as described in: https://github.com/couchbase/couchbase-lite-core/issues/437#issuecomment-370039341

Before

1. The rev tree branch that cblite considered a winner was determined as a winner was non-deterministic
2. In the case where the shorter rev tree branch was chosen as the winning branch by cblite, all subsequent updates to the cblite doc failed to push, as they were rejected by the SG proposeChanges endpoint

After

1. Both of those issues appear to be resolved
2. Still some caveats mentioned in https://github.com/couchbase/couchbase-lite-core/issues/437#issuecomment-370045740 -- this probably requires more discussion to figure out how to handle these cases.

[adamcfraser]

I don't think the non-deterministic branch choosing is really the reason we want to set sendConflicts=false. It's more correct to say here that CBL 2.0 (and blip_sync) doesn't support branched revision trees, and so we only want to send the winning revision.

[tleyden]

I guess the angle I was taking was to try to scare anyone who decided to flip that to true on a whim to fix some other bug. At least, let them know the consequences. But yeah, I agree that's not the primary motivation, so I'll reword that and leave the link to the issue for further backstory.