More debugging from (Anthropic) Claude

I asked Claude, for a second time, for a round of review on the PyQrackIsing code, for any improvements they would make. Most of Claude's suggestions did not survive developer scrutiny in discussion, but they did manage to catch a tiny bug in spin-glass cases and a bug in user-specified initial cut guesses: both of these have been fixed.

Full Changelog: v9.11.3...v9.11.4

sha1sum results:
f96307af43e0c79966c57189b15445c9e3fbd582 pyqrackising-9.11.4-py3-none-macosx_14_0_arm64.whl
4622d15d67e84b9a3672cd5666e7a65188ac43e6 pyqrackising-9.11.4-py3-none-macosx_15_0_arm64.whl
727f012e47e84490514fdaeb2cc839f4f1d5e666 pyqrackising-9.11.4-py3-none-manylinux_2_35_x86_64.whl
8a549a4cbc06ccd8a04079620e70927de33c84d7 pyqrackising-9.11.4-py3-none-manylinux_2_39_x86_64.whl
0a0fb29e0f1c736730039963b619f2567ed7d90a pyqrackising-9.11.4-py3-none-win_amd64.whl
351683885fbf3e6a4c45b6757144f782b7961a31 pyqrackising-9.11.4.tar.gz