Add tests for Matrix cast

[farzonl]

llvm/llvm-project#162711

ie support a cast like so:

export int3x2 fn_helper(float2x3 f23) {
    int3x2 i32 = {f23};
    return i32;
}

export int3x2 fn() {
    float2x3 f23 = {1,2,3,4,5,6};
    return fn_helper(f23);
}

We will need to xFail the clang side because shape transformations are not happening.
This is because the curren't indexing is wrong

https://godbolt.org/z/36bGTsfxa

+ %2 = extractelement <6 x float> %0, i32 5, !dbg !33
+ %3 = extractelement <6 x float> %0, i32 3, !dbg !33
+ %4 = extractelement <6 x float> %0, i32 1, !dbg !33
+ %5 = extractelement <6 x float> %0, i32 4, !dbg !33
+ %6 = extractelement <6 x float> %0, i32 2, !dbg !33
+ %7 = extractelement <6 x float> %0, i32 0, !dbg !33
- %2 = extractelement <6 x float> %0, i32 0, !dbg !33
- %3 = extractelement <6 x float> %0, i32 1, !dbg !33
- %4 = extractelement <6 x float> %0, i32 2, !dbg !33
- %5 = extractelement <6 x float> %0, i32 3, !dbg !33
- %6 = extractelement <6 x float> %0, i32 4, !dbg !33
- %7 = extractelement <6 x float> %0, i32 5, !dbg !33