Development roadmap

[sshin23]

Here's the development roadmap. In the short term, we should focus on:

• Performance optimization: we should ensure that factorize! and solve! doesn't allocate memory
• GPU implementation: we should make sure that we can formulate/solve problems on GPUs using the relevant BLAS/LAPACK routines in CUDA.jl
• Batched parallelism
• Nested partitioning

In the long term we should consider several generalizations/specializations:

• Portability: run on non-NVIDIA GPUs
• Specialization for dynamic programs
• Generalization for HODLR matrices
• Block multi-diagonal systems

Also, we should think of a cool name. We should convey a message that this solver targets structured sparse-dense systems

[jindavid]

• Write an interface to SparseMatrixCSC.
• Write a function to solve for multiple right-hand sides.