CUDA-accelerated ASE D3 calculator

[dambi3613]

It compiles libpaird3.so using nvcc when installing via pip. If the compilation fails, it is skipped.
If libpaird3.so is available, the D3Calculator can be used.

Could you review a few changes that might be aggressive?

1. Change 'sevennet_calculator.py' to 'calculator.py'.
2. Name the fake pair d3 implementation code for the ASE calculator to pair_d3_for_ase.cu and .h (it uses pair_d3_pars.h, which is also shared with pair_d3.cu for LAMMPS).

Additionally, I will reinforce the code related to D3 numerical tests.

[YutackPark]

We need to solve below before merge. High priority comes first.

• Fallback implementation of CPU when CUDA is not available. It is OK to slow but we need it.
• Optional d3 install (like pip install .[d3_ase]) or at least should not abort even if nvcc is not found.
• D3 unit test
• ASE SumCalculator interface (one line calculator init)

For second, currently I got this error if I didn't source cuda things:

      !!        cmd_obj.run()      CUDA is not installed or nvcc is not available.Skipping compilation of libpaird3.      error: can't copy '/tmp/tmpv1vvfpe9.build-lib/libpaird3.cpython-311-x86_64-linux-gnu.so': doesn't exist or not a regular file      [end of output]                                                                                                                                                                                                                                                                                                                                   note: This error originates from a subprocess, and is likely not a problem with pip.  ERROR: Failed building editable for sevenn                                                                                                                               Failed to build sevennERROR: ERROR: Failed to build installable wheels for some pyproject.toml based projects (sevenn)

[YutackPark]

git  clone {SevenNet}
cd SevenNet
pip install .
ls sevenn/

지금 setup.py에 '-o', 'sevenn/libpaird3.so', 로 되어있는데 이러면 패키지 인스톨 시 sevenn/ 아래에 libpaird3.so 를 만들고 site-packages에는 설치를 안함
그래서 클론한 폴더에서 테스트하면 다 찾아지는데 다른 폴더로 이동하면 못 찾아 editable install 안되는 이슈도 같은 이유같음

build_ext 에서 실제 사용하는 방식으로 하지 않고 subprocess.run 으로 구현되어 있어서 그런 것 같은데 대안이

1. build_ext 가 원래 디자인 된 대로 사용한다
2. torch 를 이용한다

2번은 torch 내장으로 유저 custom code 컴파일하고 불러오는 기능이 있는데 그걸 활용하는 거고 1번은 나도 좀 찾아봤는데 제대로 된 docs가 없는 것 같더라..