Support regular parquet in test statistics

[zigaLuksic]

WARNING: if you review this before Monday I will draw an unflattering doodle of you. Do not play games with me!

Not sure if we should have a special case where we still try to load the parquet as a geoparquet. Seems kinda hard to do in general.

[zigaLuksic]

i tested this in FB and the geometry column is problematic, not jsonifiable...

[zigaLuksic]

@mlubej any ideas how to solve the issue of WKB not being json serializable?

[zigaLuksic]

@mlubej one option i thought of was that the tester supplies the CRS. Wont work for parquets where each geometry potentially has a different CRS though...

[zigaLuksic]

tested it out a bit, the wkb->wkt also works fine. Kept the user-supplied CRS so that we can have more statistics in those cases

[mlubej]

updates (@zigaLuksic):

• added pyogrio as default gpd reading engine
• set the num_random_values to 0 for parquet stats, code block is still there and redundant
• saved also dtypes (as strings) along with columns for parquet and vector files

[mlubej]

Added also an agg_stats for dataframe-related stats and included only mean/std/min/max.

The tests for tests/pipelines/test_import_vector.py need to be updated, but we don't have the mechanism in place to update the stats via CI artifacts.

@zigaLuksic do you think it makes sense to add that also for eo-grow? I can prepare that in a separate MR if you agree, otherwise I can just manually update from a linux machine