Add HeronR2 device profile

[LakshikkaNathan]

Closes #15

Summary

Adds HeronR2() to src/profiles.jl, mirroring HeronR3(). Covers the 156-qubit heavy-hex Heron r2 processor (ibm_marrakesh, ibm_fez, ibm_kingston - one shared profile) and models the first 8 qubits as a linear chain.

Values:

• CZ: 68 ns, 2.848e-3 (published spec, arXiv:2410.00916; matches the ibm_marrakesh calibration I pulled 2026-06-08, CZ median 2.825e-3).
• SX/X: 36 ns, 3.24e-4 (ibm_marrakesh calibration). H is a synthesized estimate (SX·RZ), not native.
• T1 218 us, T2 264 us (published medians, arXiv:2410.00916).

Testing

Exports HeronR2 and adds a construction test plus the issue-requested compile_block fidelity test, leaving the existing HeronR3 construction test in place. The compile test runs at reduced size (n_levels=2, N_knots=5, max_iter=2) for speed. Full suite passes locally: 113 tests, Julia 1.12.6, Piccolo 1.18.

Note

Frequency, anharmonicity, and coupling strength are representative values. The ibm_marrakesh calibration export lists T1/T2, gate errors and durations, readout, and CZ/RZZ errors but not these three, and backend.qubit_properties() returns frequency and anharmonicity as None. So I used representative values, matching how HeronR3 is built.

AI usage

I used Claude as a coding assistant to help draft this change. I pulled the calibration myself, ran the tests, and reviewed the full patch before submitting.

[jack-champagne]

Hi @LakshikkaNathan, thanks for your submission to #15 for unitaryHACK! 👋 We've done an initial read-through and it looks like a real, well-formed contribution. We're pulling it down to run the test suite locally tonight and will follow up with detailed feedback in the morning. Appreciate your patience!

[jack-champagne]

We have many quality submissions and it was a tough decision, but there is still so much work to do on this package and we encourage everyone who made a submission to continue contributing and making suggestions, PRs, and issues on this repo. Thank you all!