Add initial support for Apple M3 and Apple M4

[alalazo]

I'm not sure this is entirely correct, but we can refine later:

% clang -dM -E - -mcpu=apple-m4 < /dev/null | grep __ARM_FEATURE_ | cut -d _ -f 5- | cut -d " " -f 1 | sort                  
AES
ATOMICS
BF16
BF16_SCALAR_ARITHMETIC
BF16_VECTOR_ARITHMETIC
BTI
CLZ
COMPLEX
CRC32
CRYPTO
DIRECTED_ROUNDING
DIV
DOTPROD
FMA
FP16_FML
FP16_SCALAR_ARITHMETIC
FP16_VECTOR_ARITHMETIC
FRINT
IDIV
JCVT
LDREX
LOCALLY_STREAMING
MATMUL_INT8
NUMERIC_MAXMIN
PAUTH
QRDMX
RCPC
SHA2
SHA3
SHA512
SME
SME2
UNALIGNED

% diff features_m2.txt features_m3.txt
% diff features_m4.txt features_m3.txt                                                                                       
22d21
< LOCALLY_STREAMING
31,32d29
< SME
< SME2