[MSLITE][Develop] fix bug of arm cpu fp16 op matmul, init const tensor before judging infershape done

mindspore/lite/nnacl/assembly/fp16/Float16ToFloat32.S

@@ -14,6 +14,8 @@ Float16ToFloat32:
     // https://github.com/ARM-software/abi-aa/blob/master/aapcs64/aapcs64.rst#simd-and-floating-point-registers
     // x19 ~ x29 should be also preserved
     // whereas our coding style do not permit such amount of parameters
+    cmp x2, #0
+    beq LoopEnd
     cmp x2, #64
     blt Loop
     Loop64:

mindspore/lite/nnacl/assembly/fp16/Float32ToFloat16.S

@@ -14,6 +14,8 @@ Float32ToFloat16:
     // https://github.com/ARM-software/abi-aa/blob/master/aapcs64/aapcs64.rst#simd-and-floating-point-registers
     // x19 ~ x29 should be also preserved
     // whereas our coding style do not permit such amount of parameters
+    cmp x2, #0
+    beq LoopEnd
     cmp x2, #64
     blt Loop
     Loop64:

mindspore/lite/src/runtime/kernel/arm/fp16/matmul_fp16.h

@@ -39,6 +39,10 @@ class MatmulFP16CPUKernel : public MatmulBaseCPUKernel {
   int RunImpl(int task_id);
 
  private:
+  int MallocMatrixABuffer();
+  int MallocMatrixBBuffer();
+  int InitBias();
+  int MallocFp16Output();
   void InitMatrixA(float *a_ptr, float16_t *a_pack_ptr);
   void InitMatrixA(float16_t *a_ptr, float16_t *a_pack_ptr);
   void InitMatrixB(float *b_ptr, float16_t *b_pack_ptr);

mindspore/lite/src/runtime/kernel/arm/fp32/transpose_fp32.cc

@@ -162,7 +162,7 @@ int TransposeCPUKernel::Run() {
     return ret;
   }
   return ret;
-}  // namespace mindspore::kernel
+}
 
 kernel::LiteKernel *CpuTransposeFp32KernelCreator(const std::vector<lite::Tensor *> &inputs,
                                                   const std::vector<lite::Tensor *> &outputs, OpParameter *opParameter,