Implement SYCL local memory

[1e-to]

We need an analog to implement the behavior numba.cuda.shared.array

• Possibly we need data structure for local memory.
• dppy.local.static_alloc(shape=10, dtype=float32) should looks like array creation dppy.local.array(shape=10, dtype=float32)
• dtype should be Numba or NumPy type? Try to change to NumPy types.

[reazulhoque]

This issue is resolved. Example of local memory creation in test here.