Fix GPU segfaults in PW USPP calculations (#7647)

* Fix GPU pointer accesses from CPU code in USPP PW calculation

* Switch radial_fft_q to template version in addusdens_g to fix linker error

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* Refactor: rename device/host pointers in USPP PW code

* minor formatting cleanup

* Add TODO comments for future USPP GPU optimizations

* style: split single-line variable declarations

* try to reduce transmission times of data

* Reduce the transmission times of data successfully

* minor formatting cleanup

---------

Co-authored-by: chengleizheng <you@example.com>
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>

Author: 3 people