fix(rust/sedona,python/sedonadb): Ensure empty batches are not included in RecordBatchReader output

[paleolimbot]

This PR ensures that if output contains batches with zero rows, they are not included in the output RecordBatchReader. Most tools can handle empty batches but not all have been tested and at least lonboard fails for this case.

Reprex from original issue:

from arro3.core import Table
import sedona.db

sd = sedona.db.connect()
sd.read_parquet("https://github.com/geoarrow/geoarrow-data/releases/download/v0.2.0/ns-water_water-poly_geo.parquet").to_view("lakes")
sd.read_parquet("https://github.com/geoarrow/geoarrow-data/releases/download/v0.2.0/ns-water_water-line_geo.parquet").to_view("rivers")
sd.sql("""SELECT geometry AS lake FROM lakes WHERE "OBJECTID" = 1976""").to_view("east_lake")

inlets_and_outlets = sd.sql("""
    SELECT "OBJECTID", "FEAT_CODE", geometry
    FROM rivers
    JOIN east_lake ON ST_Intersects(east_lake.lake, rivers.geometry)
    """)

table_orig = Table.from_arrow(inlets_and_outlets)
col = table_orig.column("geometry")
", ".join(str(len(chunk)) for chunk in col.chunks)
#> '31'

I checked R as well:

library(sedonadb)

sd_read_parquet("/Users/dewey/gh/sedona-db/submodules/geoarrow-data/ns-water/files/ns-water_water-point_geo.parquet") |> 
  sd_to_view("point", overwrite = TRUE)

sd_read_parquet("/Users/dewey/gh/sedona-db/submodules/geoarrow-data/ns-water/files/ns-water_water-junc_geo.parquet") |> 
  sd_to_view("junc", overwrite = TRUE)

sd_sql('SELECT geometry FROM junc WHERE "OBJECTID" = 1814') |> 
  sd_to_view("junc_filter", overwrite = TRUE)

joined <- sd_sql('
  SELECT "OBJECTID", "FEAT_CODE", point.geometry
  FROM point
  JOIN junc_filter ON ST_DWithin(junc_filter.geometry, point.geometry, 10000)
')

joined |> 
  nanoarrow::as_nanoarrow_array_stream() |> 
  nanoarrow::collect_array_stream() |> 
  lapply("[[", "length")
#> [[1]]
#> [1] 24

Closes #156.