Pyqrack

With this binary update, fewer Clifford ("stabilizer") edge cases result in a needless conversion to "Schrödinger method" ket (exponential cost) edge cases, particularly upon measurement across the full qubit width of a simulator.

Also, CPU simulation has been slightly further specialized in the primary single qubit gate simulation method, which therefore also benefits CPU/GPU "hybrid" methods, although perhaps imperceptibly.

This release makes a small improvement in certain cases of CPU gate simulation, and hence this benefits hybrid CPU/GPU simulation as well. Also, the "smart defaults" have been improved in Qrack's `QUnitMulti` for multi-device OpenCL simulation.

The last release failed to package Windows libraries and executables with the `none-any` wheel, and it accidentally disabled underlying qrack library environment variables. This patch fixes both issues.

This release improves CPU/GPU hybridization performance, in the underlying Qrack library. It also converts PyQrack to Qrack v7 API, but this does not affect PyQrack's shared linkage.

This fixes a problem with the simulator constructor, in v0.7.7. (Sorry for any problems that caused.)

This release adds an option to enable `QBinaryDecisionTree` as a layer in the Qrack optimization stack. (The option is off by default.) `QBinaryDecisionTree` is a CPU-based simulation layer that can significantly reduce overall RAM footprint for single simulator instances, at the cost of increasing execution time.

(This release also gives user code the control to disable the `QPager` layer. Typically, this optimization layer should be left on for systems with an available OpenCL device, as by default, or turned off for CPU-only systems.)