[FEA] Add more functionality to rolling objects

For our use cases, it would be helpful if rolling window objects have more functionality.

Basically, this would mean adding support for median, variance, standard deviation, quantile, and aggregate functions for rolling objects.

An example of support for aggregate() is below:
```
import cudf
df = cudf.DataFrame({'x': np.arange(10)})
print(df.x.rolling(2).aggregate({'x':'max'}))
```

**EDIT**
As of 7/11/2022, the following methods exist: min, mean, max, std, var, sum, count, apply. What remains is to implement:
- [ ] median (see also #6276)
- [ ] quantile
- [ ] aggregate