Pyqrack

This release add the option to round `RZ` gadget buffers that have phase angles between -Pi/8 and Pi/8 to no-operation. (Phase angles outside the range -3Pi/8 to 3Pi/8 are similarly implicitly rounded to `S` or `IS` gates.) Also, a minor bug introduced in v1.19.0 in `QUnit::GetAmplitudeOrProb()` was fixed.

It's worth noting that if you're using mirror circuit tests and validation, like the Qrack developers are, this rounding behavior leads to optimistic bias on mirror circuit benchmarks. Instead of simply using the rounded methods with mirror circuits, one might estimate fidelity from some combination of the rounded and un-rounded results of `permutation_expectation_rdm()`.

File SHA-1 sums:
caa128fbafade415bdd7ffc39124d43a1034939c pyqrack-1.19.1-py3-none-any.whl
5c035dcd3c73bb0673743480c17b2ffadc0b0ff6 pyqrack-1.19.1-py3-none-linux_aarch64.whl
31bb3729163eb305660521e34a7d521deb9b4ff2 pyqrack-1.19.1-py3-none-linux_armv7l.whl
f4fa12192eae920a00ff403805f8f1275affc720 pyqrack-1.19.1-py3-none-macosx_10_4_x86_64.whl
ea87e6b70dc038f3cb538cbe1759b7eae26f73c4 pyqrack-1.19.1-py3-none-manylinux2014_x86_64.whl
cd8d4a17147987e4b4b8a3193beed88038f342f8 pyqrack-1.19.1-py3-none-manylinux_2_35_x86_64.whl
55c1d799319cf43ee153553c454b7774b92f8298 pyqrack-1.19.1-py3-none-win32.whl
0f2107343b2f0a39c770fd8a0b6644664a19fe9d pyqrack-1.19.1-py3-none-win_amd64.whl
7a7446c67f5839f2940610ec47ed865b5a8031d6 pyqrack-1.19.1.tar.gz

**Full Changelog**: https://github.com/unitaryfund/pyqrack/compare/v1.19.0...v1.19.1

A number of "reduced density matrix" expectation value calculation methods have been added to both the static and shared library API. The point of these methods, only relevant to Clifford+RZ stabilizer simulation, is that they trace out a "reduced density matrix" excluding ancillary qubits for non-Clifford phase gadgets, instead of fully incorporating the non-Clifford phase effects. This makes these methods very fast and very low-memory, but they constitute approximations to the actual expectation values, were non-Clifford effects fully included.

File SHA-1 sums:
e2abfed0a04cc6df31894f5183f13c7c27c886ea pyqrack-1.19.0-py3-none-any.whl
64bc1be6a8e372d3989e491b70db6ad7b72cdba5 pyqrack-1.19.0-py3-none-linux_aarch64.whl
311b453026a2fc8a5ed49f21436917b09ef1aabf pyqrack-1.19.0-py3-none-linux_armv7l.whl
17fa5d478bc0764be8153e225435792cb61dfe82 pyqrack-1.19.0-py3-none-macosx_10_4_x86_64.whl
03c188eb41cd5956484e25887b074d3c1e85d725 pyqrack-1.19.0-py3-none-manylinux2014_x86_64.whl
47ae4ffaa6d8105c5994614c6382c9b3908f26fa pyqrack-1.19.0-py3-none-manylinux_2_35_x86_64.whl
a52bbf079c2a2fbc9dc2b74b822080f9990ef3b7 pyqrack-1.19.0-py3-none-win32.whl
4d1413d6182ae8a57cab92b1f24441db9af9e5a9 pyqrack-1.19.0-py3-none-win_amd64.whl
f0c7259efe25ceb7f18b037dff710dc1f59d8825 pyqrack-1.19.0.tar.gz

**Full Changelog**: https://github.com/unitaryfund/pyqrack/compare/v1.18.6...v1.19.0

This release adds the methods `clone()` and `inverse()` to the `QCircuit`/`QrackCircuit` API. (`Inverse()` produces a clone of the circuit that is the exact inverse of the original circuit.)

File SHA-1 sums:
a37aff7e3c7929fa32cd0ad7073e848c7406e76d pyqrack-1.18.6-py3-none-any.whl
5d50f08529d80dd842937bb3aa006569fe6c4e96 pyqrack-1.18.6-py3-none-linux_aarch64.whl
2ad94557caa99f0b0853f508808706eb8644edba pyqrack-1.18.6-py3-none-linux_armv7l.whl
b6e1ffce17cb9de41462c3c82d8fb24463537dd4 pyqrack-1.18.6-py3-none-macosx_10_4_x86_64.whl
f3f656ce12083a29407eed27bf8c874d6ea60a79 pyqrack-1.18.6-py3-none-manylinux2014_x86_64.whl
3516d224b44eba49f21597eec9b974583b964580 pyqrack-1.18.6-py3-none-manylinux_2_35_x86_64.whl
551aa19220be84ab3f01388deb626d83df12a47f pyqrack-1.18.6-py3-none-win32.whl
22fa26b87edc0d500d83b59f099df17c978bbd1c pyqrack-1.18.6-py3-none-win_amd64.whl
67869a136a329ec47458a0e72bfbe2cb522610e5 pyqrack-1.18.6.tar.gz

In previous patches, for "weak sampling mode," if an ancilla collapsed into |1> state, a correction was applied by replacing the injected phase gate with its inverse. This happened to work for `T` gates, but it was not correct: rather, the gate should corrected with an additional `Z` gate. This should improve weak sampling mode fidelity, again.

File SHA-1 sums:
5d758379abd9bf04ec08b75fd4d377f4d6343cc3 pyqrack-1.18.5-py3-none-any.whl
8adb4aea9ed8147b516b74c5ab50f75654543880 pyqrack-1.18.5-py3-none-linux_aarch64.whl
c6defb3fbb00d8ef0ca8e53d3fc6847fddab29cb pyqrack-1.18.5-py3-none-linux_armv7l.whl
d970293a26c8e9a03b118c1cc11f96a13ea2e488 pyqrack-1.18.5-py3-none-macosx_10_4_x86_64.whl
51c11c63d076540c1ab31fbd001e72f8ab0015c0 pyqrack-1.18.5-py3-none-manylinux2014_x86_64.whl
84b04b334800f5932531c2d2c1f45974050ce5a9 pyqrack-1.18.5-py3-none-manylinux_2_35_x86_64.whl
62f579c54fac884d5d7b0eadfdf75139f4ca8d82 pyqrack-1.18.5-py3-none-win32.whl
d541254444c2828140cb67f512e6200de55b86a4 pyqrack-1.18.5-py3-none-win_amd64.whl
eca19a563360d4a1a090fdc3fa7af72018bd287a pyqrack-1.18.5.tar.gz

All C++ Qrack unit tests currently pass for `--enable-weak-sampling` with the exception of `test_mirror_circuit` and `test_mirror_qcircuit`, which are not necessarily expected to pass due to the fact that they exceed Clifford+RZ gate set. (Qrack should automatically detect the deviation from the intended "weak sampling" gate set, in the ideal, hence this might be a target area for improvement, for next steps.)

File SHA-1 sums:
07926b8c7d0b5a06270b9c1c0546bdd79fc8369b pyqrack-1.18.4-py3-none-any.whl
d5548d45f3f68ce37d657da663a06fcfcfcf807b pyqrack-1.18.4-py3-none-linux_aarch64.whl
95321e494beee8d4f6e54071c765db9da7a1d7ce pyqrack-1.18.4-py3-none-linux_armv7l.whl
c1a37caab0561c2ec8b867684975d4da786c39ca pyqrack-1.18.4-py3-none-macosx_10_4_x86_64.whl
933d2ddfc3d173b7b40e478d758ec355e6dde85a pyqrack-1.18.4-py3-none-manylinux2014_x86_64.whl
50aa937bb78174ee30934c3f6aa67b2df43f8a7f pyqrack-1.18.4-py3-none-manylinux_2_35_x86_64.whl
0875ebcd5643c4d9653bdff2cb69387a26ac8763 pyqrack-1.18.4-py3-none-win32.whl
118d6bf66b3983eb97292e3ae3e90a451801cd2c pyqrack-1.18.4-py3-none-win_amd64.whl
76a51d8bc9a2163b43b604cbe13e122dd8ff3cc0 pyqrack-1.18.4.tar.gz

A major fidelity improvement has been made in the "weak sampling" algorithm for Clifford+RZ gate set, particularly with QUnit optimization layer, which now works with weak sampling.

File SHA-1 sums:
e9b910540ef0e1f59353b03692957c6a4dc38a9e pyqrack-1.18.3-py3-none-any.whl
548241c187d34e9208c6357989ad3f3156da5b59 pyqrack-1.18.3-py3-none-linux_aarch64.whl
5dfc59a37e40519912595cdb2168dd617feb70c9 pyqrack-1.18.3-py3-none-linux_armv7l.whl
a8ed41f96f6c97d803c639f90a87a2bac6fb20ac pyqrack-1.18.3-py3-none-macosx_10_4_x86_64.whl
c803c387dd1e8e70cb7fadaf7fd9d8a3b5ac0c56 pyqrack-1.18.3-py3-none-manylinux2014_x86_64.whl
12f0b89ed125a97e418fee1f1746f713e3825fd3 pyqrack-1.18.3-py3-none-manylinux_2_35_x86_64.whl
c1068b0cb0e44daac76bf3f404f16ca7cce65e67 pyqrack-1.18.3-py3-none-win32.whl
28377602f43f1f7256bc74514e3118b53299a72b pyqrack-1.18.3-py3-none-win_amd64.whl
57ff00a8a17b619c557d31be015a9baaf39e00b3 pyqrack-1.18.3.tar.gz