Core: Fix bin packing when target file size is smaller than a row (#2844)

## What does this change do?
When `write.target-file-size-bytes` is smaller than a single row, bin
packing computed a 0 row chunk size and PyArrow raised a ValueError.
This change clamps the chunk size to at least 1, so writes still succeed
(one row per batch/file when needed).

## Why is this needed?
Fixes a crash when users set a small target file size and attempt to
write large records.

## How was this tested?
- make lint
- uv run python -m pytest tests/io/test_pyarrow.py -k
"bin_pack_arrow_table" -v
- make test (timed out at ~42%)

Closes #2795

Co-authored-by: Soham <010Soham@users.noreply.github.com>

Author: 010Soham

pyiceberg/io/pyarrow.py

@@ -2681,7 +2681,7 @@ def bin_pack_arrow_table(tbl: pa.Table, target_file_size: int) -> Iterator[list[
     from pyiceberg.utils.bin_packing import PackingIterator
 
     avg_row_size_bytes = tbl.nbytes / tbl.num_rows
-    target_rows_per_file = target_file_size // avg_row_size_bytes
+    target_rows_per_file = max(1, int(target_file_size / avg_row_size_bytes))
     batches = tbl.to_batches(max_chunksize=target_rows_per_file)
     bin_packed_record_batches = PackingIterator(
         items=batches,

tests/io/test_pyarrow.py

@@ -2248,6 +2248,12 @@ def test_bin_pack_arrow_table(arrow_table_with_null: pa.Table) -> None:
     assert len(list(bin_packed)) == 5
 
 
+def test_bin_pack_arrow_table_target_size_smaller_than_row(arrow_table_with_null: pa.Table) -> None:
+    bin_packed = list(bin_pack_arrow_table(arrow_table_with_null, target_file_size=1))
+    assert len(bin_packed) == arrow_table_with_null.num_rows
+    assert sum(batch.num_rows for bin_ in bin_packed for batch in bin_) == arrow_table_with_null.num_rows
+
+
 def test_schema_mismatch_type(table_schema_simple: Schema) -> None:
     other_schema = pa.schema(
         (