Support for sparse data in Caffe

[beniz]

This is the integration of PR BVLC/caffe#2364 into Caffe custom build for DeepDetect.

In applications, useful for BOW models. The PR above takes data from CSR database (http://netlib.org/linalg/html_templates/node91.html#SECTION00931100000000000000), which should be automatically computed by input connector.

[beniz]

Relevant repository for NLP with Caffe, https://github.com/Russell91/nlpcaffe

[beniz]

New custom Caffe branch with support for sparse inputs is here: https://github.com/beniz/caffe/tree/master_dd_integ_sparse

[beniz]

Sparse data support for Caffe is now complete, and rebased against recent master, see comment at BVLC/caffe#2364 (comment)

This is unlikely to make it into Caffe fast (if ever), but we will maintain it against official Caffe over time.

Next step is to add sparse input support to DD's input connector (such as SVM format and special packing for NLP via txt input connector).