Restore the ability to read remote directories

[hombit]

Closes #392

[codecov]

Codecov Report

❌ Patch coverage is 69.69697% with 10 lines in your changes missing coverage. Please review.
✅ Project coverage is 97.33%. Comparing base (c0ba0f2) to head (1e15f09).
⚠️ Report is 3 commits behind head on main.

Files with missing lines	Patch %	Lines
src/nested_pandas/nestedframe/io.py	69.69%	10 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #393      +/-   ##
==========================================
- Coverage   97.79%   97.33%   -0.47%     
==========================================
  Files          19       19              
  Lines        2042     2062      +20     
==========================================
+ Hits         1997     2007      +10     
- Misses         45       55      +10     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[github-actions]

Before [960c0ec]	After [8bc53a3]	Ratio	Benchmark (Parameter)
11.1±0.3ms	11.9±0.7ms	1.07	benchmarks.NestedFrameAddNested.time_run
1.26±0.02ms	1.30±0.01ms	1.03	benchmarks.NestedFrameReduce.time_run
174M	180M	1.03	benchmarks.ReadFewColumnsHTTPS.peakmem_run
1.3G	1.34G	1.03	benchmarks.ReadFewColumnsS3.peakmem_run
258M	260M	1.01	benchmarks.ReassignHalfOfNestedSeries.peakmem_run
253M	253M	1	benchmarks.AssignSingleDfToNestedSeries.peakmem_run
134M	134M	1	benchmarks.CountNestedBy.peakmem_run
64.2±0.3ms	63.9±0.5ms	1	benchmarks.CountNestedBy.time_run
103M	103M	1	benchmarks.NestedFrameAddNested.peakmem_run
108M	107M	1	benchmarks.NestedFrameQuery.peakmem_run

Click here to view all benchmarks.

[gitosaurus]

Thanks for the fix! This looks good to me, but I have one remaining concern about whether this is supporting more than was supported before.

[gitosaurus]

Looks good!

[hombit]

Actually reading directories with column selection doesn't work right now, working on another reimplementation

[hombit]

Reimplemented again, now is_dir is called on almost every path, which, unfortunately, would cause an extra round-trip. We can try to fix it on HATS land by passing the path and filesystem separately, and adding a trailing "/" to directory paths.

It is also really poorly tested, but hopefully will be better covered by cloud tests.