Pyqrack

This release updates Qrack binaries with `TrySeparate()` improvements, for Pauli X and Y basis, and for the 2-qubit method variant.

This fixes a segmentation fault in Qrack's `QPager`.

Exception handling has been added for all expected C++ exceptions! When a C++ exception is raised in the Qrack library binaries, it will be caught, and a Python exception will be automatically thrown instead.

Also, user code now has the option to directly disable OpenCL when constructing a simulator.

Significant improvements have been made in internal Schmidt decomposition performance optimization and optimal default stack stability, in the underlying Qrack binary, (particularly when using the environment variable `QRACK_QUNIT_SEPARABILITY_THRESHOLD`).

When using the `export QRACK_QUNIT_SEPARABILITY_THRESHOLD=0.25` environment variable for a tradeoff between accuracy and efficiency, for example, (which accepts a value between `0` and `0.5`) it has become apparent that "paging" with the `isPaging=True` kwarg can be unstable in a way that interferes with this important special case of simulation. Also, the default value of this kwarg doesn't necessarily make sense: with default options, paging **does** engage when single GPU segment maximum allocation is exceeded, even though paging remains **off** for any lower number of qubits. Hence `isPaging=True` is probably a better way to describe this same original case, while the ability to force paging off with `isPaging=False` can give greater stability and efficiency to approximate simulation in excess of ~30 qubits on a laptop or desktop computer.

The vm6502q.7.0.0 Qrack binaries remain completely unchanged, from v0.8.6.

This release has exact parity with the official tagging and release of the underlying Qrack v7 API.

- Asynchronous parallelism has been slightly reduced, with no appreciable decrease in performance, which helps prevent "Resource temporarily unavailable" exceptions.
- Underlying separability checks in `TrySeparate()` have been improved from discrete X/Y/Z checks to continuous inverse single qubit state preparation checks, which theoretically greatly improves the chances of success of separability checks when single qubits are actually separable, (but nonspecific circuits will probably not see a noticeable reduction in resource usage).