[Ruby] Provide an equivalent of Pyarrow's .to_pylist

[fpacanowski]

Describe the enhancement requested

In certain use cases it's quite useful to convert a RecordBatch or a Table to an array of hashes, e.g. when using Parquet for data exchange. Pyarrow provides this functionality as to_pylist function.

Basically I'd love to see an inverse of Arrow::RecordBatchBuilder.build.

rb = Arrow::RecordBatchBuilder.build(schema, data)
assert rb.to_pylist == data

Additionally it would be great if it correctly handled struct and list data type, for example:

schema = Arrow::Schema.new([
  Arrow::Field.new("c1", :int32),
  Arrow::Field.new("c2",
    Arrow::StructDataType.new([
      Arrow::Field.new("c3", :int32),
      Arrow::Field.new("c4",
        Arrow::StructDataType.new([
          Arrow::Field.new("c5", :int32),
          Arrow::Field.new("c6", Arrow::ListDataType.new(:int32)),
        ])
      )
    ])
  )
])

data = [
    {c1: 1, c2: {c3: 3, c4: {c5: 5, c6: [6, 7]}}},
    {c1: 11, c2: {c3: 13, c4: {c5: 15, c6: [16, 17]}}},
]
rb = Arrow::RecordBatchBuilder.build(schema, data)
assert rb.to_pylist == data

(Obviously the method wouldn't be named to_pylist.)

Component(s)

Ruby

[kou]

We have Arrow::RecordBatch#raw_records/Arrow::Table#raw_records and #each_raw_record.
But they don't add keys: [1, {"c3" => 3, "c4" => {"c5" => 5, "c6" => [6, 7]}}]