[compiler-rt] [compiler-rt] Don't provide `__arm_sme_state` for baremetal targets (PR #191434)

[Benjamin Maxwell via llvm-commits]

================
@@ -69,7 +69,9 @@ struct {
 
 // The formatter wants to re-order these includes, but doing so is incorrect:
 // clang-format off
-#if defined(__APPLE__)
+#if defined(TARGET_BAREMETAL_AARCH64)
----------------
MacDue wrote:

It was to make the logic consistent between `sme-abi.S` and `aarch64.c`. So if `TARGET_BAREMETAL_AARCH64` is defined we always use the baremetal FMV and remove the `__arm_sme_state` implementation. 

https://github.com/llvm/llvm-project/pull/191434