Make cellsToLinkedMultiPolygon tests order independent

[ajfriend]

To prepare for #1127, this PR updates the cellsToLinkedMultiPolygon tests to not depend on a specific polygon order, since the cellsToMultiPolygon backend will return a different order.

Those tests should now pass before and after the migration. Also, added a test containing cells that fail for the current algo, but should pass with the new one.

[coveralls]

coverage: 99.095%. remained the same
when pulling 9ebad97 on ajfriend:aj/poly_tests
into 8ac991e on uber:master.

[ajfriend]

Note: the CLI tests in tests/cli/cellsToMultiPolygon.txt still hardcode exact vertex order. The migration will change the starting vertex within each loop (a cyclic rotation), so the expected outputs will need to be regenerated. (But when viewing the diff, large chunks should remain unchanged.)

I think just making that change with the migration is easier than trying to generalize the CLI tests. cc: @dfellis