Update auto_rebase_pass to take optional allow_swaps argument

[sjdilkes]

If allow_swaps is true, then when targeting CX or ZZMax gates, auto_rebase_pass will decompose TK2 gates with wire swaps added if it will reduce the number of CX/ZZMax gates in the decomposition.

To do this I've added two new methods to CircPool.hpp and the library.cpp binders: TK2_using_CX_and_SWAP and TK2_using_ZZMax_and_SWAP.

These methods construct circuits via a new method in CircPool.cpp TK2_Swap_replacement, which finds cases where wire swaps improve gate count.

To cleanup remaining single qubit gates, I've also updated standard_rebase_via_tk2 to move the TK2 gate decomposition into the for loop that replaces all gates with TK2 operations. This makes it easier apply single qubit cleanup to just the gates as part of that decomposition, shouldn't effect performance of the overall method, and seemed to be necessary to make gates such as SWAP return only a wire swap and not any further single qubit gates.

[cqc-alec]

Nice, just a few comments.

[cqc-alec]

A couple of pytket tests failing?

[sjdilkes]

A couple of pytket tests failing?

I forgot to update the n_gates value for Sycamore for ZZPhase and ZZMax, have fixed now

[cqc-alec]

A couple of pytket tests failing?

I forgot to update the n_gates value for Sycamore for ZZPhase and ZZMax, have fixed now

Just the one still failing...

[cqc-alec]

thanks!