Local SWAP Fusion via `applyMultiSwap()

[VinaYrdx]

Submission: Local SWAP Fusion via applyMultiSwap()

I have implemented the local (non-distributed) SWAP fusion described in this issue.

Fork: https://github.com/VinaYrdx/QuEST
Branch: main
Key commit: e7a3d71

What was implemented

• New public API function applyMultiSwap(Qureg, vector<int>, vector<int>) in operations.h + operations.cpp
• O(k) qubit permutation via disjoint transposition composition in cpu_statevec_multiSwap_fused_sub()
• Single O(2^n) pass with involution guard (j > i) — no visited array needed
• Full 3-layer dispatch: API → localiser → accelerator → CPU subroutine
• Graceful fallback to sequential applySwap() for distributed/GPU quregs

Benchmark (n=24, single-threaded, OpenMP disabled)

k (pairs)	Sequential	Fused	Speedup
3	0.229s	0.097s	2.36x
5	0.458s	0.163s	2.82x
8	0.433s	0.214s	2.02x

This complements PR #786's distributed case — together they cover both local and distributed SWAP fusion.

[VinaYrdx]

Submitted PR #795 implementing local (non-distributed) SWAP fusion via applyMultiSwap().

Benchmark at n=24 (single-threaded): 2.36x–2.82x speedup over sequential SWAPs. Full details in the PR.