Add GPU Cholesky Decomposition

[rok-cesnovar]

Description

Adds the GPU Cholesky decomposition for the matrix_gpu type, and adds the GPU functionalites to the /prim and /rev implementations of the Cholesky Decomposition. This issue will be tackled with 2 pull requests:

• the kernel code for small blocks, the cholesky_decompose(matrix_gpu&) function and /prim implementation
• /rev implementation

Example

Use of cholesky_decompose(matrix_gpu&):

stan::math::matrix_gpu m0(m0_eigen);
using stan::math::cholesky_decompose;
stan::math::matrix_gpu foo = cholesky_decompose(m0);

Use of cholesky_decompose(stan::math::matrix_d) and cholesky_decompose(stan::math::matrix_v):

using stan::math::cholesky_decompose;
stan::math::matrix_gpu foo = cholesky_decompose(m0_eigen);

Expected Output

The expected output is faster execution of Cholesky Decomposition.

Additional Information

/

Current Math Version

v2.18.0