[XPU] Fix the topk, set_value ops that using temporary tensors avoiding the memory overlaps during multi-stream inference (#54851)

* add unique kernel for xpu

* add unique kernel for xpu

* update uniittest

* add xpu support for unique with axis

[XPU][PHI Kernels] bind bitwise_add kernel & add int32/int64 support to scatter_nd_add kernel for xpu (#54066)

* bind new kernels to xpu

* refine code

* fix bugs in unittest

* add pad op for xpu

* add pad op for xpu

* add pad op for xpu

* add_depthwise_conv2d_transpose

* Update test_depthwise_conv2d_transpose_op_xpu.py

删除print语句

* fix as review, add fp16 conv2d_transpose

* fix unittest of bn and reduce_mean

* fix bn unittest

* fix ci

* fix ci

* bind sparse_coo_tensor, reduce_max/max_int32, range/arange_int32, equal_bool, scatter_grad_float32, nearest_interp_int64 kernels

* add more unit tests; modify compilation logic of xpu sparse kernels

* clang format

* add cumsum_grad op to xpu2_op_list

* [XPU] add numel op

* [XPU] update numel/size op registration

instance_norm_grad kernel. Fix bugs that the data type of input is different from output in reduce_sum kernel. test=kunlun

* [XPU] add fp16 support for compare ops.

* fix ci.

* Delete randperm raw op

* fix

* [XPU] add increment op.

* fix ci

* [XPU] add reduce_min_grad XPU kernel

* add unittest for reduce_min xpu op

* [xpu-ernie3] support fp16 for full_like/stack/where xpu kernels

* [xpu-ernie3] support fp16 for full_like/stack/where xpu kernels