Fix compilation errors related to cube_direction operators

[eltank]

Summary

Bugfixes "Fix compilation errors related to cube_direction operators"

Purpose of change

Band-aid fix for recent compilation issue affecting most PRs. See #51317

Describe the solution

Remove the stuff that's not constexpr from the constexpr operators of cube_direction.
There's probably a better way to fix this, but this should work in the short-term.

Describe alternatives you've considered

Making the operators not constexpr, dropping the static asserts and making a unit test. Requires moving the cube_direction declaration into the header.

Testing

Relying on CI to verify this.

Additional context

[jbytheway]

Thanks for catching this when I failed to notice it in CI.

This is a reasonable stopgap solution. I'll try to look for a better one tomorrow.

[actual-nh]

Once it passes the cross-compile, I'd say it's ready for merging.

[actual-nh]

WTF? It fails on clang for MacOS, precisely the compiler (and OS, possibly due to compiler) it wasn't failing earlier?

[eltank]

WTF? It fails on clang for MacOS, precisely the compiler (and OS, possibly due to compiler) it wasn't failing earlier?

The error is:

ERROR : src/overmap.cpp:1574 [std::vector<tripoint_om_omt> mutable_overmap_special_data::place(overmap &, const tripoint_om_omt &) const] Spawn of mutable special anthill had unresolved joins.  Existing terrain at (157,0,-1) was ants_ew; joins were south: tunnel_to_tunnel1.385 s: mx_minefield theoretical spawn

That's unrelated to this PR and probably non-deterministic (something for @jbytheway to look into)

[jbytheway]

Agreed, that's unrelated, non-deterministic, and for me to look into.