Prefer extracting C++ standard version from argv at startup over runtime evaluation

[anutosh491]

Currently, the C++ standard version is determined by evaluating a small C++ snippet dynamically at runtime via Cpp::Evaluate, using a function like get_stdopt():

int __get_cxx_version () {
#if __cplusplus > 202302L
    return 26;
...
#endif
}
__get_cxx_version()

While determines the version correctly, this consumes the first incremental_module and I think is also costlier especially for wasm kernels (where the user has to wait for this to execute before running a cell)

[anutosh491]

Since we already pass the C++ standard version explicitly in kernel.json via flags like the following

"argv": [
  "xcpp", "-std=c++20", ...
]

We should probably save all this effort and handle it just like xeus-cling does https://github.com/jupyter-xeus/xeus-cling/blob/93d61b98518c494fc0e195844c392fc0d40dd0d2/src/xinterpreter.cpp#L464-L479

This makes sense as the standard version enters the clang cc1 command from kernel.json file itself.

[vgvassilev]

The solution is incorrect. Please revert and discuss these things with me.