Feature: MFOV layer-prealign

[minnerbe]

I tentatively added a pre-alignment routine that treats a single layer of a single MFOV as a tile and aligns these using a rigid model.

I tested this on a small stack (see MultiSemPreAlignClientTest). So far, it seems like pulling and processing all the data is fine (and does reduce the error by some factor), but pushing the data to the render database again doesn't seem to work. In particular, the put request that should set the stack state to complete doesn't return. Do you have any clues as to why this might be, @trautmane?

Once these issues are resolved, I would test this on a larger stack and also check if this reduces the time for the subsequent alignment. Finally, I'll make this into a spark client that can process multiple stacks at once.

Side note: To put the transforms computed by this method into the TileSpecs, I wanted to use SolveTools.getTransformSpec, but the TODOs there seem to indicate that this method doesn't work right now. The last commit that changed this method says that the issue is fixed, though. Can you please comment on that, @StephanPreibisch? Resolved by 0df486.

[minnerbe]

I tested this pre-align method on this stack and got these results.

[trautmane]

My main concern is about pulling all tile and match data for each stack in a single request. Let's discuss other options.