Make subchunkable usable by people who don't want to divide

[supersergiy]

There's 3 types of bbox expansion:

1. Resolution based expansion -- bbox given by the user does not evenly divide by the dst_resolution.
2. Bunk based expansion -- bbox given by the user doesn't align with bunk of the dst layer.
3. Punk based expansion -- .... with top level punk .... This is implemented as expand_bbox rn, and it moves bottom right corner.

Proposal:
Make each one of them a boolean parameter, apply those parameters independently in 1->2->3 order.

When both (2) and (3) are set, require some sort of divisibility between top level punk and dst.bunk.

[supersergiy]

@dodamih What do you think about the last possible mode:

Do (1) and (3), not (2), and allow non-alligned writes during the final write, either everywhere or at the boundaries only

[dodamih]

It's not difficult in principle, but we have enough case checking as is at the moment which will only increase with the PR that will fix this - I think allowing non-aligned writes at the top level should be a separate issue that we can deal with once we have docs.

Subchunkable is going to have 25 total possible arguments with the latest integration test PR and the expand + direct write PR - we're at a point where it's difficult for me to keep track of and reason through how all the args affecting BBox interact with each other.