fix(parquet/pqarrow): Fix propagation of field-ids for Lists (#397)

### Rationale for this change
An issue was found in apache/iceberg-go when attempting to retrieve data
from a table containing a List Column that had a struct as the element.
It was failing to propagate the element-id for the fields when fetching.
I tracked it down to the schema handling here.

### What changes are included in this PR?
Changes the `getNestedFactory` method in pqarrow/schema.go to use
`ListOfField` instead of `ListOf` so that it preserves the metadata,
i.e. the field id.

### Are these changes tested?
Yes, a test has been added to cover this scenario.

### Are there any user-facing changes?
Previously this situation would result in a field-id of -1, now users
will see the field-id get propagated correctly.

Author: zeroshade

parquet/pqarrow/encode_dictionary_test.go

@@ -688,7 +688,9 @@ func TestArrowWriteNestedSubfieldDictionary(t *testing.T) {
 	dictValues := array.NewDictionaryArray(dictType, indices, dict)
 	defer dictValues.Release()
 
-	data := array.NewData(arrow.ListOf(dictType), 3, []*memory.Buffer{nil, offsets.Data().Buffers()[1]},
+	data := array.NewData(arrow.ListOfField(arrow.Field{
+		Name: "element", Type: dictType, Nullable: true,
+		Metadata: arrow.NewMetadata([]string{"PARQUET:field_id"}, []string{"-1"})}), 3, []*memory.Buffer{nil, offsets.Data().Buffers()[1]},
 		[]arrow.ArrayData{dictValues.Data()}, 0, 0)
 	defer data.Release()
 	values := array.NewListData(data)

parquet/pqarrow/schema.go

@@ -1015,19 +1015,19 @@ func getNestedFactory(origin, inferred arrow.DataType) func(fieldList []arrow.Fi
 		switch origin.ID() {
 		case arrow.LIST:
 			return func(list []arrow.Field) arrow.DataType {
-				return arrow.ListOf(list[0].Type)
+				return arrow.ListOfField(list[0])
 			}
 		case arrow.FIXED_SIZE_LIST:
 			sz := origin.(*arrow.FixedSizeListType).Len()
 			return func(list []arrow.Field) arrow.DataType {
-				return arrow.FixedSizeListOf(sz, list[0].Type)
+				return arrow.FixedSizeListOfField(sz, list[0])
 			}
 		}
 	case arrow.MAP:
 		if origin.ID() == arrow.MAP {
 			return func(list []arrow.Field) arrow.DataType {
 				valType := list[0].Type.(*arrow.StructType)
-				return arrow.MapOf(valType.Field(0).Type, valType.Field(1).Type)
+				return arrow.MapOfFields(valType.Field(0), valType.Field(1))
 			}
 		}
 	}

parquet/pqarrow/schema_test.go

@@ -473,6 +473,36 @@ func TestProperListElementNullability(t *testing.T) {
 	assert.True(t, arrSchema.Equal(outSchema), "expected: %s, got: %s", arrSchema, outSchema)
 }
 
+func TestFieldNestedPropagate(t *testing.T) {
+	arrSchema := arrow.NewSchema([]arrow.Field{
+		{Name: "transformations", Type: arrow.ListOfField(
+			arrow.Field{
+				Name: "element",
+				Type: arrow.StructOf(
+					arrow.Field{Name: "destination", Type: arrow.BinaryTypes.String,
+						Metadata: arrow.NewMetadata([]string{"PARQUET:field_id"}, []string{"6"})},
+					arrow.Field{Name: "transform_type", Type: arrow.BinaryTypes.String,
+						Metadata: arrow.NewMetadata([]string{"PARQUET:field_id"}, []string{"7"})},
+					arrow.Field{Name: "transform_value", Type: arrow.BinaryTypes.String,
+						Metadata: arrow.NewMetadata([]string{"PARQUET:field_id"}, []string{"8"})},
+					arrow.Field{Name: "source_cols", Type: arrow.ListOfField(
+						arrow.Field{Name: "element", Type: arrow.BinaryTypes.String, Metadata: arrow.NewMetadata([]string{"PARQUET:field_id"}, []string{"10"})}),
+						Metadata: arrow.NewMetadata([]string{"PARQUET:field_id"}, []string{"9"})},
+				),
+				Metadata: arrow.NewMetadata([]string{"PARQUET:field_id"}, []string{"5"}),
+			},
+		), Metadata: arrow.NewMetadata([]string{"PARQUET:field_id"}, []string{"4"})},
+	}, nil)
+
+	pqSchema, err := pqarrow.ToParquet(arrSchema, nil, pqarrow.DefaultWriterProps())
+	require.NoError(t, err)
+
+	result, err := pqarrow.FromParquet(pqSchema, nil, metadata.KeyValueMetadata{})
+	require.NoError(t, err)
+
+	assert.True(t, arrSchema.Equal(result), "expected: %s, got: %s", arrSchema, result)
+}
+
 func TestConvertSchemaParquetVariant(t *testing.T) {
 	// unshredded variant:
 	// optional group variant_col {