Remove pyarrow dependency

[kylebarron]

For #581

[kylebarron]

After a lot of work here and in arro3, we're pretty close to having this working! I think just three tasks!

• table rechunking so that each Parquet buffer can have reasonably similar sizes

  lonboard/lonboard/_serialization.py

  Line 34 in 4370186

  for record_batch in table.to_batches(max_chunksize=max_chunksize):

• Support for adding an array or a chunked array in table.add_column

  lonboard/lonboard/_viz.py

  Line 453 in a60adbc

  table = table.append_column("row_index", arange_col)

• inferring the data type for primitive data types in from_numpy

  lonboard/lonboard/traits.py

  Lines 1008 to 1013 in ac5baac

  	return fixed_size_list_array(
  	Array.from_numpy(
  	value.ravel("C"),
  	),
  	list_size,
  	)

[kylebarron]

This is now working!

[kylebarron]

All the tests now pass (locally)! arro3 isn't yet published to pypi, so that's why CI here is failing

[kylebarron]

I ran through a couple examples manually and it's still working as expected!