Add dynamical decoupling benchmarks on ibm_fez

[anulum]

DD sequences (XY4, CPMG) are implemented in mitigation/dd.py and runner.transpile_with_dd() but never benchmarked on hardware.

Goal: Run kuramoto_4osc at depth 85 and 149 with and without DD. Compare R(t) error.

Expected outcome: DD should reduce idle-qubit decoherence in the 85-149 depth regime where gate errors are not yet dominant.

Budget: ~2 min QPU (4 circuits x 10k shots x 2 conditions).

Blocked on: March QPU budget (available 2026-03-01).

[github-actions]

This issue has been inactive for 60 days. It will be closed in 14 days unless there is new activity. Add the pinned label to keep it open.