Draft: Improve multicore performance

Changes:

• More multicore friendly data layout, more suitable for multicore workloads.
• Integration of the TrigDx library for high-performance trigonometric computations. Not all operations are yet performed by a TrigDx lookup table, especially as some sinusoidal operations are done in double precision. Switching to floating point in these cases cause a significant drop in accuracy (>1.0e-4). The number of elements in the lookup table may be increased, or these operations should fall back to a 'compute-based' sin/cos implementation.
• ...