Proposal: combine wave correctness and convergence test pairs into a single binary

[AstroKriel]

There currently exist several wave test problems where both a correctness test (FastWave, SlowWave, AlfvenWaveLinear, HydroWave) and a Richardson convergence test (FastWaveConvergence, SlowWaveConvergence, etc.) exist, each reimplementing the same IC logic. This inflates the test count and upkeep cost, which came up in the review of #1910.

Proposed design

Consolidate each pair into the ...Convergence binary. The binary reads mode flags and parameters from a [setup] table:

Mode flags (at least one must be true):

• setup.run_convergence (default: true): runs the Richardson sweep
• setup.run_sim (default: false): runs a single fixed-resolution correctness check at the TOML-specified resolution and stop time

Convergence parameters (run_convergence = true):

• setup.nx_start — starting resolution
• setup.nx_max — ending resolution
• setup.machine_precision_target — 0 disables early exit and runs all levels to nx_max
• setup.refine_n_dims (default: 1) — dimensions to refine; see Proposal: extend Richardson convergence harness to support N-dimensional refinement #1958

Correctness parameters (run_sim = true):

• setup.error_tol — L1 error tolerance for pass/fail

Wave parameters (MHD problems only):

• setup.angle_between_k_b0 (default: 0.0)
• setup.num_modes_x

With default TOML files unchanged, CTest behaviour is identical to today.

Affected binaries

Keep	Remove
FastWaveConvergence	FastWave
SlowWaveConvergence	SlowWave
AlfvenWaveLinearConvergence	AlfvenWaveLinear
HydroWaveConvergence	HydroWave
EntropyWaveConvergence	(no paired binary; gains the param surface for consistency)

Motivation

• Eliminates redundant IC implementations across paired binaries.
• The fixed-resolution sim mode is useful for stability sweeps above the Richardson cap (e.g. probing scheme instability; see The Quokka2026 EMF scheme shows instabilities for slow waves at high resolution #1912).
• Establishes a consistent setup.* parameter surface across all wave convergence tests (currently inconsistent: HydroWaveConvergence is fully hardcoded; EntropyWaveConvergence is missing nx_start, nx_max, machine_precision_target).

Deferred

• setup.num_modes_y, setup.num_modes_z, and 2D/3D TOML configs: pending the Richardson N-dimensional refinement extension (Proposal: extend Richardson convergence harness to support N-dimensional refinement #1958).
• mhd.resistivity support in run_sim mode beyond AlfvenWaveLinear: deferred.

Relation to other proposals

Depends on the Richardson N-dimensional refinement extension (#1958).

[AstroKriel]

Resolved by #1978