- implemented ScalarQuantizer (without IVF)
- implemented update for IndexIVFFlat
- implemented L2 normalization preproc

* moved most FAISS_ASSERT calls to C++ exceptions, and adjusted
  memory allocation to avoid mem leaks

* added an IndexIVFScalarQuantizer type that offers an
  intermediate compression between IVFFlat and IVFPQ

* support removal of indices in IndexIDMap / IndexFlat combination

* various fixes in GPU code

  memory allocation to avoid mem leaks

* added an IndexIVFScalarQuantizer type that offers an
  intermediate compression between IVFFlat and IVFPQ

* support removal of indices in IndexIDMap / IndexFlat combination

* various fixes in GPU code

The rootdir in the distribution seems specific to the creators.  I just made the rootdir for the fvecs be the suggested directory (sift1M) from the installation guide.

correct utils::nextHighestPowerOf2

The bit shift was done as int and not as current type (size_t for example), thus the computed next highest power of 2 was wrong on large numbers.

* add initial cmake support

* update cmake, add cmake instructions to INSTALL

* update findopenmp and INSTALL

* change FindOpenBLAS.cmake to cater for macports

- change cblas.h to openblas_config.h since macports does not ship
cblas.h with openblas.

* revise INSTALL for cmake

Added additional notice concerning compute capabilities.

* Docker build from local sources rather than Github cloned sources

* Update INSTALL instructions for docker build from Github repo

* Back to generic makefile.inc.Linux and use of env var in Dockerfile

* Back to generic makefile.inc.Linux and use of env var in Dockerfile

avoid reading beyond the end of an array in fvec_L2sqr and related functions

* Add Dockerfile

* Update documentation with Docker build/run details

- better selection of training sets for PQ and preprocessing
- GPU parameter object
- IndexIDMap fixed
- fixed redo bug in clustering

- better selection of training sets for PQ and preprocessing
- GPU parameter object
- IndexIDMap fixed
- fixed redo bug in clustering

swiggfaiss_wrap.cxx already include <numpy/arrayobject.h>, if compiled with -I/usr/lib64/python2.7/site-packages/numpy/core/include/numpy, file will be searched at /usr/lib64/python2.7/site-packages/numpy/core/include/numpy/numpy/arrayobject.h and will not be found.

* Changing a few settings for MKL on Ubuntu

This is my current configuration for MKL on Ubuntu.
I also corrected the `BLASLDFLAGS=-DFINTEGER=long` line (it should be `BLASCFLAGS`).

* Update makefile.inc.Linux

Tested and documented on MLK 2017.0.098