Including test

GPU Matrix Algebra and Operators

On windows use smaller batchsize

Merge branch 'gpu_basic_matrix_algebra' of https://github.com/bstatcomp/math into gpu_basic_matrix_algebra

Added vectorized version of bernoulli_logit_rng (Issue 45)

Increase runTests.py batchsize to make use of more cores.

Fix arguments of sub block so any section of the source block can go in any source of the destination block

  - set_kernel_args is a helper function that passes in kernel arguments
   using variadic templating.
  - sub_block places a submatrix of a source matrix into another matrix.
    The subset always starts at position (0,0) in the destination matrix.

set_kernel_args added to the gpu code.

This reverts commit 2442c067.

Merge branch 'feature/issue-45-vectorized-bernoulli-logit' of https://github.com/stan-dev/math into feature/issue-45-vectorized-bernoulli-logit