Multiple sync-related fixes

[almet]

I found some issues with the current implementation of the sync mechanism, which this PR fixes.

• The "move a feature from one datalayer to another" feature was not thought of, at all, and as a result introduced the bug that was referenced in #1931.
• It turns out that the syncing of this operation wasn't working either, creating shape clones along the way.
• The previous (buggy) implementation was not sending the proper layer id, resulting to features hosted by the wrong layers.
• Operations were triggered when loading data from the original server, which shouldn't happen.

Generally speaking, the current implementation misses some cross-datalayer tests, but I suggest these should come as a separate pull request.

[almet]

Rebased on latest master, and incorporated the proposed changed from reviews. We should be ready to go.