Provide a ByteBuffer based write path

Similar to the effort to use a ByteBuffer in the read path, we should provide a path to use a ByteBuffer in the write path.

The proposed idea is to provide an allocator in the writer that is passed down to CapacityByteArrayOutputStream and is used to allocate a ByteBuffer instead of the byte array that it currently uses.

I've attached a patch based on the 1.5 release that does that. There are two implementations of the allocator - one for Heap memory and one for Direct memory. In addition, the allocator interface allows for a 'release' method so that reference counted memory allocators like Netty's ByteBuf can be used to provide an allocator.

**Reporter**: [Parth Chandra](https://issues.apache.org/jira/secure/ViewProfile.jspa?name=parthc) / @parthchandra
#### Related issues:
- [Move off of Parquet fork](https://issues.apache.org/jira/browse/DRILL-1410) (is depended upon by)

_{**Note**: *This issue was originally created as [PARQUET-78](https://issues.apache.org/jira/browse/PARQUET-78). Please see the [migration documentation](https://issues.apache.org/jira/browse/PARQUET-2502) for further details.*}

Beginnen Sie damit, den bestehenden auf ByteBuffer basierenden Lesepfad mit dem Writer und dem in diesem Issue genannten CapacityByteArrayOutputStream zu vergleichen. Verfolgen Sie, wie der vorgeschlagene allocator Heap- oder Direct-Buffer bereitstellen würde und wie deren Freigabe behandelt würde, und überprüfen Sie anschließend, ob der Schreibpfad beide allocator-Implementierungen und referenzgezählten Speicher unterstützt.

Vom Indexierungsmodell aus dem Issue-Text verfasst.