`gravity_vlct_stable_Jeans_wave` test fails with Clang Compiler

[mabruzzo]

James reported this failure while reviewing PR #319.

The failure of that test is really weird! The only thing I could possibly imagine (since there shouldn't be any floating-point optimizations) is that default round-off behavior for floating-point values could hypothetically change.

[mabruzzo]

Some updates: this is DEFINITELY still an issue.

• I explicitly confirmed that the test passes when compiled with gcc (on macOS and on linux)
• the test fails when compiled with clang on linux and apple-clang on macOS.

I need to actually plot the simulations results, but somethings seems to be going VERY wrong. The L1 Errror norms are absolutely massive.

• My speculation is that I'm invoking some undefined behavior that happens to give nice results with gcc, but gives weird results with clang... It might be worth changing the optimization level to see how that affects things (or enable sanitizers)