LoadStoreOp can have scalar inputs, so type check is required (#3359)

Type error was detected with #3263 while I was testing it with a Debug
build.

```
pytest -v tests/python/test_python_frontend.py -k test_pad_dynamic
```

It has a fusion of:

```
Inputs:
  T0_g_float[ bS0{1}, iS1{i1}, iS2{i2} ]
Outputs:
  T1_g_float[ bS11{1}, iS12{( ( i1 + ( (nvfuser_index_t)(( (int64_t)(( float(2.5) * ( (float)(7) ) )) )) ) ) + ( (nvfuser_index_t)(( (int64_t)(( float(2.5) * ( (float)(7) ) )) )) ) )}, iS13{( ( i2 + ( (nvfuser_index_t)(( (int64_t)(( float(2.5) * ( (float)(7) ) )) )) ) ) + ( (nvfuser_index_t)(( (int64_t)(( float(2.5) * ( (float)(7) ) )) )) ) )} ]

%kernel_math {
f7 = (float)(7);
f9 = float(2.5) * f7;
i11 = (int64_t)(f9);
i14 = (nvfuser_index_t)(i11);
i16 = (nvfuser_index_t)(i11);
i18 = (nvfuser_index_t)(i11);
i20 = (nvfuser_index_t)(i11);
T2_l_float[ bS3{1}, iS5{( ( i1 + ( (nvfuser_index_t)(( (int64_t)(( float(2.5) * ( (float)(7) ) )) )) ) ) + ( (nvfuser_index_t)(( (int64_t)(( float(2.5) * ( (float)(7) ) )) )) ) )}rf, iS7{( ( i2 + ( (nvfuser_index_t)(( (int64_t)(( float(2.5) * ( (float)(7) ) )) )) ) ) + ( (nvfuser_index_t)(( (int64_t)(( float(2.5) * ( (float)(7) ) )) )) ) )}rf ]
   = pad( T0_g_float[ bS0{1}, iS1{i1}, iS2{i2} ], {0, 0, i14, i16, i18, i20} )
T1_g_float[ bS11{1}, iS12{( ( i1 + ( (nvfuser_index_t)(( (int64_t)(( float(2.5) * ( (float)(7) ) )) )) ) ) + ( (nvfuser_index_t)(( (int64_t)(( float(2.5) * ( (float)(7) ) )) )) ) )}, iS13{( ( i2 + ( (nvfuser_index_t)(( (int64_t)(( float(2.5) * ( (float)(7) ) )) )) ) ) + ( (nvfuser_index_t)(( (int64_t)(( float(2.5) * ( (float)(7) ) )) )) ) )} ]
   = Set( T2_l_float[ bS3{1}, iS5{( ( i1 + ( (nvfuser_index_t)(( (int64_t)(( float(2.5) * ( (float)(7) ) )) )) ) ) + ( (nvfuser_index_t)(( (int64_t)(( float(2.5) * ( (float)(7) ) )) )) ) )}rf, iS7{( ( i2 + ( (nvfuser_index_t)(( (int64_t)(( float(2.5) * ( (float)(7) ) )) )) ) ) + ( (nvfuser_index_t)(( (int64_t)(( float(2.5) * ( (float)(7) ) )) )) ) )}rf ], cache_op=Streaming )
} // %kernel_math
```

Stack trace:
```
#0  __cxxabiv1::__cxa_throw (obj=0xabc6ea0, tinfo=0x7ffeba81c248 <typeinfo for nvfuser::nvfError>, dest=0x7ffeba059370 <nvfuser::nvfError::~nvfError()>) at ../../../../libstdc++-v3/libsupc++/eh_throw.cc:80
#1  0x00007ffeba058665 in nvfuser::nvfCheckFail (func=0x7ffeb9927c33 "as", file=0x7ffeb99d9a1b "/raid/nmaruyama/debug1/csrc/utils.h", line=119,
    msg=0x7ffeb99c36e6 " INTERNAL ASSERT FAILED at \"/raid/nmaruyama/debug1/csrc/utils.h\":119, please report a bug with repro script to NVFuser at https://github.com/NVIDIA/Fuser/issues. ") at /raid/nmaruyama/debug1/csrc/exceptions.cpp:283
#2  0x00007ffeb9c1be4b in nvfuser::nvfErrorFail (func=0x7ffeb9927c33 "as", file=0x7ffeb99d9a1b "/raid/nmaruyama/debug1/csrc/utils.h", line=119,
    condMsg=0x7ffeb99c36e6 " INTERNAL ASSERT FAILED at \"/raid/nmaruyama/debug1/csrc/utils.h\":119, please report a bug with repro script to NVFuser at https://github.com/NVIDIA/Fuser/issues. ") at /raid/nmaruyama/debug1/csrc/exceptions.h:229
#3  0x00007ffeb9c1bbe4 in nvfuser::PolymorphicBase::as<nvfuser::TensorView> (this=0xac07490) at /raid/nmaruyama/debug1/csrc/utils.h:119
#4  0x00007ffeba54c67f in nvfuser::(anonymous namespace)::isLoadGlobalToLocal (expr=0xabd7c50) at /raid/nmaruyama/debug1/csrc/scheduler/cache_policy_refiner.cpp:61
#5  0x00007ffeba54c599 in nvfuser::refineCachePolicy (fusion=0xabef940) at /raid/nmaruyama/debug1/csrc/scheduler/cache_policy_refiner.cpp:153
```

Author: naoyam

csrc/scheduler/cache_policy_refiner.cpp

@@ -58,6 +58,12 @@ bool isLoadGlobalToLocal(const Expr* expr) {
   if (ldst->opType() != LoadStoreOpType::Set) {
     return false;
   }
+  // It should not be necessary to check the output since it should be
+  // always a TensorView as long as the input is a TensorView, but
+  // just in case.
+  if (!ldst->in()->isA<TensorView>() || !ldst->out()->isA<TensorView>()) {
+    return false;
+  }
   if (ldst->in()->as<TensorView>()->getMemoryType() != MemoryType::Global) {
     return false;
   }