It seems GCC supports _Float16 in C++ on aarch64 since GCC 13, at least
according to my quick tests on godbolt.org.

(It was already supported in C since GCC 7.)