Refactor top-level interface

[Jefffrey]

(Where top-level interface refers to how DataFusion will use this library to read ORC files as that is the main intention of the crate)

Since we want this library to integrate with DataFusion, we should try provide a more clean interface for it to be able to read ORC files as record batches.

In current way:

datafusion-orc/tests/basic/main.rs

Lines 14 to 22 in cbeb8bd

	fn new_arrow_reader(path: &str, fields: &[&str]) -> ArrowReader<File> {
	let f = File::open(path).expect("no file found");
	let reader = Reader::new(f).unwrap();
	let cursor = Cursor::new(reader, fields).unwrap();
	ArrowReader::new(cursor, None)
	}

• Create reader (our struct) from file
• Create cursor (our struct) from reader
• Create arrow reader (our struct) from cursor

Similar can be said for async version.

We can take inspiration from how parquet does it:

• https://docs.rs/parquet/latest/parquet/arrow/arrow_reader/struct.ParquetRecordBatchReader.html
• https://docs.rs/parquet/latest/parquet/arrow/arrow_reader/type.ParquetRecordBatchReaderBuilder.html
• https://docs.rs/parquet/latest/parquet/arrow/async_reader/struct.ParquetRecordBatchStream.html
• https://docs.rs/parquet/latest/parquet/arrow/async_reader/type.ParquetRecordBatchStreamBuilder.html

[Jefffrey]

I will work on trying to simplify the Reader/Cursor part a bit, maybe try to replicate what parquet does here: https://github.com/apache/arrow-rs/blob/master/parquet/src/file/reader.rs#L40-L68