[AArch64][SVE] Redundant svand_z with all-true masks is not optimized unlike GCC

[k-arrows]

Test case:

#include <arm_sve.h>

svbool_t test (svbool_t pg, svbool_t prev)
{
  return svand_z (svptrue_b8 (),
                  svpnext_b16 (prev, pg),
                  svptrue_b16 ());
}

https://godbolt.org/z/Pr14zMsWa

[llvmbot]

@llvm/issue-subscribers-backend-aarch64

Author: None (k-arrows)

Test case: ```cpp #include <arm_sve.h>

svbool_t test (svbool_t pg, svbool_t prev)
{
return svand_z (svptrue_b8 (),
svpnext_b16 (prev, pg),
svptrue_b16 ());
}

https://godbolt.org/z/Pr14zMsWa
</details>