Consolidate examples and test data

[tschaub]

We have a single example Parquet file, a couple JSON metadata files, and an R file in the examples directory. In addition, we have a number of examples Parquet files in the test_data directory. We should consolidate these (likely remove examples and point people to test_data for examples).

All our example Parquet files should be generated by scripts in the repo. And if we need to have separate JSON metadata files, these should be generated too (though it feels like we could do without these). And after we consolidate the examples into a single location, the release process should be updated to describe how to update the examples for new versions.

[jorisvandenbossche]

While it would indeed be good to consolidate things, the current example.parquet is more of an actual (useful) example in the sense that it contains real-world data. While the test_data files just contains some dummy coordinates and no real attributes. So there might be some value in keeping an example file (or some) as well?