__cpp_lib_atomic_float
The corresponding feature was implemented in LLVM 18 (by #67799), but this FTM wasn't added before.