Accelerate surface Green function calculation and I/O for large Lead principal layers or high-k case

[AsymmetryChou]

The current surface Green function calculation becomes a bottleneck when the Lead principal layer is relatively large. This significantly affects the overall performance and scalability of the code in large or complex systems.

Proposed enhancement:

• Investigate and implement algorithms or optimizations to accelerate the surface Green function calculation, especially for cases with large principal layers.
• Possible directions include parallelization, or leveraging efficient numerical libraries.
• Benchmark and compare performance before and after enhancement.

This improvement is important for extending the applicability of dpnegf to larger systems and for reducing computation time in practical simulations.

[AsymmetryChou]

For some cases, such as graphene as electrodes, very high k-point mesh along the transverse direction is necessary. However, the strategy now is to calculate and save all self-energies in one .h5 file once for subsequent calculations.

Therefore, dense I/O operations occur. There might be better strategy here.