Improve tests and behaviour in Summaries for anti-meridian crossing datasets

[whatnick]

Current test for anti-meridian crossing dataset summaries uses only a single polygon. This does not really exercise the database side ST_UNION function. In practice a small number of datasets can be crafted where the union results in degenerate summary polygons. See the illustrations below.

Datasets:

Product Summary:

Relevant source:

datacube-explorer/cubedash/summary/_summarise.py

Line 72 in ccf1f3a

select_by_srid = (

Relevant test:

datacube-explorer/integration_tests/test_model.py

Line 170 in ccf1f3a

def test_footprint_antimeridian(benchmark):