Python wheel builds are failing with handler timeout

[paleolimbot]

After #1067 and #1051 we have better hygene with our context management and Python GDAL usage; however, we still have failing tests:

Windows fails after the Python UDF tests, and Linux still intermittently fails after the GDAL/OGR read via Python.

https://github.com/apache/sedona-db/actions/runs/29952109746/job/89051129963#step:5:63075

https://github.com/apache/sedona-db/actions/runs/29952109746/job/89032091791#step:4:79348

It's very possible that #1071 can fix this (fewer or static lifetimes), plus moving away from pyogrio as the source of OGR reads.

[paleolimbot]

For Windows it's coming back to me that I once had a hang in the UDF tests where the read path went through pyogrio. In this case there was too much Python GIL interaction (I think) or some other lock and there was a deadlock where the query wouldn't finish.