Implement non-nested struct statistic extraction

[Lordworms]

Which issue does this PR close?

Closes #10609

Rationale for this change

What changes are included in this PR?

Are these changes tested?

Are there any user-facing changes?

[Lordworms]

Did not implement the nested struct type since the current RowGroupMetaData and ColumnChunkMetaData seems not to support nesting

[alamb]

perhaps we can update the comments here too

[Lordworms]

Got it

[alamb]

Thank you @Lordworms

[alamb]

@tustvold could you help here -- is this the correct way to find the correct parquet leaf from an arrow struct type?

[alamb]

Something isn't right here -- the structs themselves aren't nullable (only some of their child fields)

datafusion/datafusion/core/tests/parquet/mod.rs

Lines 954 to 958 in 7c08a6f

	let struct_array_data = struct_array(vec![
	(Some(1), Some(6.0), Some(12.0)),
	(Some(2), Some(8.5), None),
	(None, Some(8.5), Some(14.0)),
	]);

datafusion/datafusion/core/tests/parquet/mod.rs

Lines 1024 to 1045 in 7c08a6f

	fn struct_array(input: Vec<(Option<i32>, Option<f32>, Option<f64>)>) -> ArrayRef {
	let int_32: Int32Array = input.iter().map(|(i, _, _)| i).collect();
	let float_32: Float32Array = input.iter().map(|(_, f, _)| f).collect();
	let float_64: Float64Array = input.iter().map(|(_, _, f)| f).collect();
	let nullable = true;
	let struct_array = StructArray::from(vec![
	(
	Arc::new(Field::new("int32_col", DataType::Int32, nullable)),
	Arc::new(int_32) as ArrayRef,
	),
	(
	Arc::new(Field::new("float32_col", DataType::Float32, nullable)),
	Arc::new(float_32) as ArrayRef,
	),
	(
	Arc::new(Field::new("float64_col", DataType::Float64, nullable)),
	Arc::new(float_64) as ArrayRef,
	),
	]);
	Arc::new(struct_array)
	}

[alamb]

I think nulls only have to check the current nulls (not their children's nulls)...

[Lordworms]

sure, I'll fix it