Using pandas groupby but keeping the order of other columns, then shift within group

Question

Say I have a pandas dataframe df as shown below -

    a   b
0   1   23
1   2   67
2   1   98
3   1   45
4   2   64
5   3   76

I