Python-zpar

- Accompanying release for [ZPar v0.7.5](https://github.com/frcchang/zpar/releases/tag/v0.7.5) which is a big bugfix release.
- Fixed segfaults when using python-zpar interactively.
- Removed hacky fix for single word sentences introduced in `v0.9.2` since the underlying bug has been fixed in ZPar.
- Previously we were programmatically redirecting STDOUT to STDERR because ZPar used to print informational messages to STDOUT. However, this has been fixed in the new release of ZPar. This redirection is no longer necessary and has been removed.

Set default logging level to be `WARNING` instead of `INFO`, for `Tagger.py`, `Parser.py`, and `DepParser.py`.

The latest version of ZPar has a bug where it produces non-deterministic output for sentences that contain a single word in all caps. This hack title-cases such words to make the output deterministic and then restores the original word. This hack will be removed once the underlying bug in ZPar is fixed which is under progress.

- Adding the new functions to parse tagged data to the server script.
- Updating the client script to show examples of these new functions.

- This release adds functions called `[dep_]parse_tagged_sent()` and `[dep_]parse_tagged_file()` that allow the user to obtain constituency and dependency parses for already tagged sentences and files.
- It also adds simple unit tests for all the major functions.

- ZPar development has now moved entirely to GitHub, from SourceForge. The [ZPar homepage](http://people.sutd.edu.sg/~yue_zhang/doc/index.html) will be updated to reflect this change soon.
- ZPar English and Chinese models will be attached to a release if they have been updated. If not, the release notes will link to the old models. For this release, only the tagger model was updated (see below) but since this is the first release on GitHub, all models are linked.
- Fixed null pointer bugs such that the constituency parser should no longer crash on empty lines (2).
- Fix an underflow bug that was causing inconsistent tagging behavior for single word sentences (4). This required retraining the tagger model.
- The Spanish ZPar combined binary (`zpar.es`) now compiles successfully.
- Modify code to print out informational/logging messages to STDERR instead of STDOUT.
- Fix typo `outout` -> `output` in entire codebase.
- Other minor fixes and cleanup.