Add gpu kernel for new api : linalg.lstsq · Pull Request !2720 · PaddlePaddle/Paddle

### PR types

Function optimization
### PR changes

OPs
### Describe

Add gpu kernel for new api : linalg.lstsq
You can see the implementation of cpu kernel in [PR38585](https://github.com/PaddlePaddle/Paddle/pull/38585)
The docs_cn of paddle.linalg.lstsq is merged by [docs-PR4174](https://github.com/PaddlePaddle/docs/pull/4174)

1、usage example：
```
import paddle
paddle.set_device("cpu")
x = paddle.to_tensor([[1, 3], [3, 2], [5, 6.]])
y = paddle.to_tensor([[3, 4, 6], [5, 3, 4], [1, 2, 1.]])
results = paddle.linalg.lstsq(x, y, driver="gelsd")

# solution
print(results[0]) 
# [[ 0.78350395, -0.22165027, -0.62371236],
# [-0.11340097,  0.78866047,  1.14948535]]

# residuals
print(results[1])
# [19.81443405, 10.43814468, 30.56185532])

# rank
print(results[2])
# 2

# singular values
print(results[3])
# [9.03455734, 1.54167950]
```

2、docs_cn：

![image](https://user-images.githubusercontent.com/86215757/148715214-c250cd16-c7a2-4916-af15-312bcd3e8925.png)

3、docs_en：

![image](https://user-images.githubusercontent.com/86215757/148715244-3a7b7ed5-424f-4c8f-bd6f-de599e8e8a54.png)

GVPPaddlePaddle/Paddle

搜索帮助

GVP PaddlePaddle/Paddle