test(ibm): add a golden for the Vieille's-law pressure-coupled burn rate

The ibm_burning_grain/ibm_flameholder Example goldens already cover surface injection (v_blow + the inj_species composition swap), but both leave burn_rate_exp at 0, so the pressure-coupled scaling v_blow*(p/burn_rate_pref)^n shipped with no regression coverage. Registers the same example with --burn_exp 0.5 at the Example suite's 25x25/50-step size, so it differs from the existing golden only by the coupling: turning it on moves the solution by ~5e-3 relative (1.3 absolute in energy), well above the 1e-3 tolerance, so a broken multiplier fails the test. Tolerance pinned to 1e-3 to match the sibling Example golden -- compute_tolerance would otherwise pick 1e-10 via the ib branch, which a 50-step IBM + finite-rate-chemistry run cannot hold across compilers.

Author: sbryngelson