Q# should include decompositions to alternate quantum gate sets

[swernli]

Q# defines the set of basic gate operations in the Microsoft.Quantum.Intrinsic namespace, where the fundamental operations are implemented as body intrinsic and other operations are expressed as decompositions onto these body-intrinsic operations. These effectively represent a specific quantum gate set that the current version of the simulator implements. To facilitate experimentation, testing, and eventually targeting of simulators and hardware with a different quantum gate set, the runtime should have infrastructure and examples of alternate implementations of operations in the Microsoft.Quantum.Intrinsic namespace as decompositions on top of the body-intrinsic operations for that alternate gate set.

Introducing this functionality will require several tasks.

• [x] Refactor Qsharp.Core into the foundational elements shared by all decompositions and the replaceable elements that are override by alternate decompositions. (Resolved by #254)

• [x] Introduce the first new decomposition package on top of a different quantum gate set with alternate implementations for Microsoft.Quantum.Intrinsic operations. (Resolved by #363)

• [x] Allow alternate decompositions to be tested against the default QuantumSimulator to verify behavior. (Resolved by #363)

• [x] Refactor the EntryPointDriver to allow for target specific simulators. (Moved out to issue #388)

• [x] Fix e2e build to allow integration of refactored QsharpCore into main (#397, #414)

• [x] Add decomposition packages for other common quantum gate sets. (Resolved by #367)

• [x] Add documentation on individual decomposition packages and general guidance on authoring new packages.

• [x] Rework use of internal in defining callabes that can be implemented outside of their original binary (see discussion in #367). (Moved out to issue #389)

[bamarsha]

@swernli Can this issue be closed or is there still work in progress?

[swernli]

@samarsha You beat me by one minute :). The new PR I opened (#476) is the last one needed to finally resolve this issue. It's the big merge of the feature branch into main, and will likely take a while, but should auto-close this once it is done.