x86 sse2 for loongarch: fix GCC build failure

[jinboson]

It seems there is some problem for GCC in dealing with vector type on Ternary Operator, see below:

//test.c
#define SIMDE_ENABLE_NATIVE_ALIASES
#include "/home/jin/data/community/simde-loong64/simde/x86/sse2.h"
int main()
{
        __m128i a = {8};
        __m128i b = _mm_slli_si128(a, 3);
        return 0;
}

gcc -mlsx test.c，the compile output is:

In file included from test.c:2:
test.c: In function ‘main’:
/home/jin/data/community/simde-loong64/simde/x86/sse2.h:1347:22: error: type mismatch in conditional expression
   (((imm8)<=0) ? (a) : (((imm8)>15) ? simde_mm_setzero_si128() : simde__m128i_from_lsx_i8((v16i8)__lsx_vbsll_v(simde__m128i_to_private(a).lsx_i64, (imm8)))))
                      ^
/home/jin/data/community/simde-loong64/simde/x86/sse2.h:1406:35: note: in expansion of macro ‘simde_mm_bslli_si128’
   #define _mm_slli_si128(a, imm8) simde_mm_bslli_si128(a, imm8)
                                   ^~~~~~~~~~~~~~~~~~~~
test.c:6:14: note: in expansion of macro ‘_mm_slli_si128’
  __m128i b = _mm_slli_si128(a, 3);
              ^~~~~~~~~~~~~~

We can simplify the demo like bellow, which is tested on x86:

//1.c
#include <immintrin.h>
int main() {
        __m128i a, b;
        int i = 3;
        b = i > 0 ? a : a + a;
        return 0;
}

gcc 1.c, the compile output is:

> 
> > 1.c: In function ‘main’:
> > 1.c:5:23: error: type mismatch in conditional expression
> >     5 |         b = i > 0 ? a : a + a;
> >       |                       ^

I have reported the strange phenomenon to our compiler team, they are wroking on it.
But I think there is no harm in adding the simde__m128i type-cast for (a) to work round this problem.

[jinboson]

Reproducing this failure requires a and b to be declared by __m128i, which is why the CI test didn't catch this before. So it's not the problem of CI.

[jinboson]

@mr-c ping.