Another attempt at introducing a columnated merge batcher

[antiguru]

This builds on #414 but uses the new Layout feature recently merged.

[frankmcsherry]

I have another question too: could the merger be for arbitrary BatchContainer types, or does it just work for TimelyStack? Is there some BC::prefered_capacity() method we might add to make it general?

[antiguru]

I have another question too: could the merger be for arbitrary BatchContainer types, or does it just work for TimelyStack? Is there some BC::prefered_capacity() method we might add to make it general?

I'd prefer a single general implementation, but it's not easy to get there:

• For the vector-based implementation, we can recycle buffers from the input and don't need to traverse the pending buffer. We could, but we first would need to do the measurements to see what impact it might have.
• For Columnation-based containers, we have a nice copy_destructured API, which allows us to avoid allocating a tuple just because we have no alternative way to represent its parts. We don't have this API for Vec, so we'd need to invent something.
• The rest seems to be mostly typing work, so not very difficult, and it might create some clarity around the C parameter to OrdValBatch.

[frankmcsherry]

Looks good, and let's merge!