feat: Implement _mm_aesdeclast_si128

[howjmay]

closes #477

[wangxiao1254]

I got a compiler error here
https://github.com/emp-toolkit/emp-tool/actions/runs/3784574277/jobs/6433980013#step:4:36

and fixed by

@@ -9889,8 +9889,8 @@ FORCE_INLINE __m128i _mm_aesenclast_si128(__m128i a, __m128i RoundKey)
 FORCE_INLINE __m128i _mm_aesdeclast_si128(__m128i a, __m128i RoundKey)
 {
     return vreinterpretq_m128i_u8(
-               vaesdq_u8(vreinterpretq_u8_m128i(a), vdupq_n_u8(0))) ^
-           vreinterpretq_u8_m128i(RoundKey);
+               vaesdq_u8(vreinterpretq_u8_m128i(a), vdupq_n_u8(0)) ^
+           vreinterpretq_u8_m128i(RoundKey));
 }

[jserv]

Thank @wangxiao1254 for reporting! Applied in commit 6fd46b6 .

@howjmay, we have to automate the cryptographic extension aware tests.

[wangxiao1254]

I think the ) that you deleted in line 9892 should be moved to line 9893 :) I think this should fix #572 as well.

Also the function at line 9898 should be updated as well. Currently it would cause type mismatch when calling vaesimcq_u8 with a in type int64x2_t.

FORCE_INLINE __m128i _mm_aesimc_si128(__m128i a)
 {
-    return vreinterpretq_m128i_u8(vaesimcq_u8(vreinterpretq_u8_m128i(a)));
+    return vreinterpretq_m128i_u8(vaesimcq_u8(a));
 }