我有以下数据框：
df
session logtime speed
0 49 1458203631 0.020396
1 49 1458203632 13.049023
2 49 1458203633 0.068780
3 49 1458203634 NaN
4 58 1458216259 6.729906
5 58 1458216260 6.155009
6 58 1458216261 29.300476
7 58 1458216262 NaN
每个会话的速度的最后一个值是NaN。我想用以前的速度值替换这些NaN，如下所示：
session logtime speed
0 49 1458203631 0.020396
1 49 1458203632 13.049023
2 49 1458203633 0.068780
3 49 1458203634 0.068780
4 58 1458216259 6.729906
5 58 1458216260 6.155009
6 58 1458216261 29.300476
7 58 1458216262 29.300476