Parquet

This release contains a number of fixes and enhancements since version 1.1.

**Fixes**:

* Support nulls (optional) values when the column is encoding using the PLAIN_DICTIONARY encoding. 37
* Fix parsing of int96 49
* Fix IndexError for some bitpacked data 60

**Improvements**:

* Provider better error message when file is opened in the wrong mode 33
* Implement logical-types: Automatically translate from raw to logical types like dates,
unsigned ints, utf-8, etc.
* pep8, pylint, and pydoc fixes/improvements.

This release contains a number of fixes and enhancements since version 1.0. The main features of this release are Python 3 support and improved performance. See below for more details of the release.

**Fixes:**
- Support for definition levels (reading optional fields with nulls). 28
- README updates to better capture state of project and missing features.

**Improvements:**
- Simplified `read_rle` implementation.
- Optimized code-paths with debug logging.
- Bulk reads for plain encoding to drastically improve performance (as much as 10x) 26.
- Switches to thriftpy library (for supporting python 3).
- new, pythonic API (see README) 11.