[v3] create integration test suite comparing zarr-python 2 and 3

[jhamman]

We could use a set of integration tests that compare zarr-python 2 and 3. Topics for these tests may include:

1. API compatibility (can the same code be run w/ 2.x and 3.x?)
2. Specification consistency (do both versions produces spec consistent data?)
3. Interoperability (can data written with 2.x be read with 3.x and visa versa?)

Developing this ahead of the 3.x release will help communicate where we have gaps and where we have confidence.

xref: #2182 (comment)

[jhamman]

@dstansby - I think this can be moved back to after 3.0.0. For (1), we have the Xarray and Dask test suites now running on Zarr 2 and 3 so we're feeling confident there. More can certainly be done but this doesn't have to be done before 3.0.0.