Added CI fix

From: @yuchaojie
Reviewed-by: 
Signed-off-by:

From: @caojian05
Reviewed-by: @wuxuejian,@zhouyaqiang0,@kisnwang
Signed-off-by: @wuxuejian,@kisnwang

From: @sishuikang
Reviewed-by: @zhanghaibo5,@zhang_xue_tong
Signed-off-by: @zhanghaibo5

From: @jojobugfree
Reviewed-by: @kisnwang,@zhoufeng54
Signed-off-by: @zhoufeng54

From: @jjfeing
Reviewed-by: @kisnwang,@zhoufeng54
Signed-off-by: @zhoufeng54

From: @liu_xiao_93
Reviewed-by: @liangchenghui,@wuxuejian
Signed-off-by: @liangchenghui

From: @chen_ying113
Reviewed-by: 
Signed-off-by:

From: @wuxuejian
Reviewed-by: @chujinjin,@liangchenghui
Signed-off-by: @liangchenghui

From: @corleoneliu
Reviewed-by: @ckey_dou
Signed-off-by:

From: @penny369
Reviewed-by: @wuxuejian,@oacjiewen,@c_34
Signed-off-by: @c_34

From: @zyhstack
Reviewed-by: @ouwenchang,@ouwenchang,@ouwenchang,@wangyue01
Signed-off-by: @wangyue01

From: @liu_xiao_93
Reviewed-by: @liangchenghui,@chujinjin
Signed-off-by: @liangchenghui

From: @gaoxiong1
Reviewed-by: @dylangeng,@anyrenwei
Signed-off-by:

From: @dayschan
Reviewed-by: @gaoxiong1,@ckey_dou
Signed-off-by: @ckey_dou

From: @lianliguang
Reviewed-by: @zh_qh
Signed-off-by: @zh_qh

From: @zhoufeng54
Reviewed-by: @kisnwang,@chujinjin
Signed-off-by: @chujinjin

Signed-off-by: zhoufeng <zhoufeng54@huawei.com>

From: @hangangqiang
Reviewed-by: @zhang_xue_tong,@zhanghaibo5
Signed-off-by: @zhang_xue_tong

From: @zhang_yi2020
Reviewed-by: @gemini524,@wuxuejian,@liangchenghui
Signed-off-by: @wuxuejian

1. synchronized the commit 8af78cd5c(Added ExpandDims into GPU fusion list),
   but reverted the modification in basic_ops_fusion.cc(8af78cd5c)
2. added ExpandDims into ShapeOpsSplitter's process list, and moved this pass to before GraphKernelSplitter,
   changed it to process sub func_graph only.

From: @nicholas_yhr
Reviewed-by: @kisnwang,@zhoufeng54
Signed-off-by: @zhoufeng54

From: @c_34
Reviewed-by: @wuxuejian,@zhouyaqiang0,@oacjiewen
Signed-off-by: @oacjiewen

From: @chujinjin
Reviewed-by: @kisnwang,@zhoufeng54
Signed-off-by: @zhoufeng54

From: @zhoufeng54
Reviewed-by: @kisnwang,@xu-yfei
Signed-off-by: @xu-yfei

From: @nicholas_yhr
Reviewed-by: @zhoufeng54,@chujinjin
Signed-off-by: @chujinjin

From: @dayschan
Reviewed-by: @gaoxiong1,@ckey_dou
Signed-off-by: @ckey_dou

Cast the float16-input to float32 before ReduceSum, and cast back to float16 after ReduceSum.
If the op after this ReduceSum is a casting from float16 to float32, then it can be eliminated.

From: @zjun3021
Reviewed-by: @kisnwang,@chujinjin
Signed-off-by: @chujinjin

From: @ttudu
Reviewed-by: @c_34,@oacjiewen
Signed-off-by: @c_34

From: @yuzhenhua666
Reviewed-by: @oacjiewen,@c_34
Signed-off-by: @c_34

From: @lihongkang1
Reviewed-by: @liangchenghui,@wuxuejian
Signed-off-by: @liangchenghui