Compressible module: boundary faces emit a spurious wave in 9.2.0

With an identical setup.xml and no user routine, the compressible module (constant gamma) gives a correct Sod shock tube in 9.1.0 and a corrupted one in 9.2.0. Both runs are pure first order upwind, all six boundaries are symmetry, and the time step is fixed and identical, so nothing on the input side differs between the two runs.

The corruption is entirely localised at the boundaries. The rarefaction, the contact discontinuity and the shock are unchanged, and agree with the exact Riemann solution in both versions. What 9.2.0 adds is a spurious wave emitted by each wall at the first time step, which then travels into the domain at the local speed of sound (see figure below).

The case is here, if you want to run it as is: [Sod tube](https://github.com/simvia-tech/tutorials-code_saturne/tree/main/30_compressible/Comp_Sod_Tube)

Run the linked Sod tube case from tutorials-code_saturne/30_compressible/Comp_Sod_Tube with the identical setup.xml under 9.1.0 and 9.2.0, then compare the first-step boundary behavior. Trace the compressible module's symmetry-boundary handling; done means the 9.2.0 run no longer emits the spurious wall waves while retaining the correct Sod solution.

Written by the indexing model from the issue text.