Should we switch over to a newer C++ version?

[crisluengo]

I found myself wanting to use inline static constexpr class member variables, but couldn't because we're using C++14, it was introduced in C++17.

What are the downsides of upgrading? Would we continuously upgrade to the latest version of the standard, or the 5-year-old version, or...? Would we upgrade only if a new feature makes it worthwhile? Should we never upgrade?

New stuff that would be useful in DIPlib:

• C++17:
  • std::filesystem::
  • std::string_view
  • std::optional and std::variant
  • if constexpr
  • inline variables (inline static)
  • structured bindings (auto [a, b, c] = ...)
  • [[fallthrough]] (currently using compiler-dependent comments)
  • [[nodiscard]] (which we're currently already using if the compiler supports it)
• C++20:
  • designated initializers ({.x = 1, .y = 2})
  • abbreviated function template (void func(auto);)
  • Concepts (simplifies template specializations)
  • std::format()
  • std::string::starts_with() and std::string::ends_with()
  • std::ranges::
• C++23:
  • std::float16
  • z and uz suffix for std::size_t and std::ptrdiff_t
  • std::flat_set, std::flat_multiset, std::flat_map, std::flat_multimap (sets and maps with fewer mallocs)