Fix ndf.add_nested() for empty frames

[hombit]

Fixes #88.

This changes the implementation of NestedFrame.add_nested to use df[name] = data instead of df.assign(name=data), because the least doesn't work for empty frames. The reason we care is dask metadata

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 98.72%. Comparing base (41cce93) to head (d4eb8fc).
Report is 72 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main      #92   +/-   ##
=======================================
  Coverage   98.71%   98.72%           
=======================================
  Files          15       15           
  Lines         859      861    +2     
=======================================
+ Hits          848      850    +2     
  Misses         11       11           

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

[github-actions]

Before [41cce93]	After [7d3ddf9]	Ratio	Benchmark (Parameter)
55.9±3ms	60.7±4ms	1.09	benchmarks.ReassignHalfOfNestedSeries.time_run
255M	257M	1.01	benchmarks.AssignSingleDfToNestedSeries.peakmem_run
272M	274M	1.01	benchmarks.ReassignHalfOfNestedSeries.peakmem_run
31.1±2ms	31.1±0.7ms	1	benchmarks.AssignSingleDfToNestedSeries.time_run
86.5M	86.6M	1	benchmarks.NestedFrameAddNested.peakmem_run
89.5M	89.4M	1	benchmarks.NestedFrameQuery.peakmem_run
89.7M	89.4M	1	benchmarks.NestedFrameReduce.peakmem_run
8.98±0.1ms	8.86±0.1ms	0.99	benchmarks.NestedFrameAddNested.time_run
5.30±0.08ms	5.17±0.09ms	0.98	benchmarks.NestedFrameReduce.time_run
6.30±0.05ms	6.11±0.02ms	0.97	benchmarks.NestedFrameQuery.time_run

Click here to view all benchmarks.