O(n) per Gray code search iteration

Gray code search iterations on GPU have been updated to calculate cut or energy differential in O(n) time. Also, a bug has been fixed in initially "seeding" distinct Gray code search threads.

We also explored CPU-based cut/energy differential calculation in O(n) time: this doesn't appear to be particularly fruitful. Cut/energy calculation on CPU is already highly optimized and sub-O(n^2) in practice: there might be no easy further gain, here.

Full Changelog: v9.7.6...v9.7.7

sha1sum results:
da3b28744f747c4cd5616f0e1e9ae5cc40c958f6 pyqrackising-9.7.7-py3-none-macosx_14_0_arm64.whl
74d7e79caa3a0e35ad9b0221db5152a3256736dc pyqrackising-9.7.7-py3-none-macosx_15_0_arm64.whl
18a59388d653b845e992b4900301e35d3e1657d3 pyqrackising-9.7.7-py3-none-manylinux_2_35_x86_64.whl
b7c3996372c855e1d828728377865a79ab57b9b0 pyqrackising-9.7.7-py3-none-manylinux_2_39_x86_64.whl
8ccde196832bcf7568f608de09ad4d82108fd70d pyqrackising-9.7.7-py3-none-win_amd64.whl
15f72b38077a61770687d1d0cc11d41d50892691 pyqrackising-9.7.7.tar.gz