Pandas reverse rolling sum using offset with varying windows sizes

Question

Consider the following dataframe:

    date                    values
0   2020-01-07 09:00:00     5
1   2020-01-07 10:00:00     8
2   2020-01-07 11:00:00     9