Genomeworks

* Bug fix : Updated `cmake` default installation path mechanism

Release 0.4.3 is a hotfix for wheel package generation in Puyclaragenomics.

* Python 3.5 was Added to PyPI classifiers.
* Fixed `liblogging.so` linkage error occurred in Python 3.5.

Release 0.4.0 adds several performance and API updates to the Clara Genomics Analysis SDK.

The following are highlights of this release -
1. **CUDA Mapper**
* Significant performance improvements to all-vs-all overlap generation using `cudamapper`.

2. **CUDA POA**
* _**New**_ C++ and Python graph APIs for inspecting graphs generated during partial order alignment process.
* Bug fixes and general improvements.

3. **CUDA Aligner**
* Major bug fix for Myers Hirschberg implementation.

4. **Pyclaragenomics**
* Improvements to `evaluate_paf` script for comparing overlap files.

5. **SDK Improvements**
* _**New**_ PyPI pcakages available for `pyclaragenomics`. Details in `pyclaragenomics` README.

Release 0.3.0 adds new tools, functionality and APIs to the Clara Genomics Analysis SDK.

The following are highlights of this release -
1. **CUDA Mapper**
_**New**_ A CUDA-accelerated implementation of the [minimap2](https://academic.oup.com/bioinformatics/article/34/18/3094/4994778) algorithm. This first release of cudamapper focuses primarily on single GPU all-to-all mapping.

2. *CUDA Aligner**
* _**New**_ Improved performance of Myers-Hirschberg implementation on batches with very long sequences (~64k)
* _**New**_ Python API for CUDA Aligner

3. *SDK Improvements*
* _**New**_ Debian and RPM packaging support for the SDK.
* _**New**_ Online [Doxygen documentation](https://clara-genomics.github.io/ClaraGenomicsAnalysis/cpp/index.html) for C++ API

The following are the major additions to this release -
1. **CUDAPOA**
* **_New_** C++ API for creating CUDAPOA batches.
* **_New_** Python API bindings for CUDAPOA in pyclaragenomics
* **_New_** Python and C++ samples to guide usage of relevant APIs
* Lower memory footprint batches, requiring less than 50% of earlier versions, allowing more POA groups to be computed within a batch.
* Overall batched POA time down by ~20% on medium to large size batches (300-400 POA groups per batch).

2. **CUDAALIGNER**
* **_New_** Myers + Hirschberg implementation for global alignment, allowing for larger batches of longer sequences to be aligned on the GPU.

3. **SDK Improvements**
* **_New_** CUDA 9.0 support added to SDK

This release is the first official release of ClaraGenomicsAnalysis SDK. The modules developed in this release are geared primarily towards long-read use cases, such as with data from Oxford Nanopore Technologies or PacBio.

The main components of this release are -

1. cudapoa - CUDA accelerated Partial Order Alignment, including support for consensus and multi-sequence alignment generation.
2. cudaaligner - CUDA accelerated batched global alignment based on Ukkonen's algorithm.