Why would they ever remove said instruction, it seem to be a very small set of extra one.
Even if we do not use any apple library, the compilers that target Apple Silicon could use those extra instruction, no ?
They could remove it because they can provide a guarantee that it won't be problematic, as it's already abstracted away from mortals using whatever library that took advantage of it under the covers.
As long as it's considered undocumented, a compiler will generally never emit it, even if you're deliberately targeting the exact architecture. It kinda causes havoc on multiple levels - imagine you're looking at the disassembly as you look at optimizing something and something is just unexplainably wrong. There's just "garbage" data that the debugger decodes to nonsense - now you have one very confused programmer who is wondering if they're looking at a compiler bug because there's no reference for whatever they're encountering.
If you happen to know the opcode, you could do it with inline assembly, but there's probably not even a mnemonic for it. You'll wind up needing to poke the literal bytes representing the instructions to force the assembler's hand. This might even be what their library does behind the scenes.