[FEA] Expanding window functions in libcudf/cuDF

**Is your feature request related to a problem? Please describe.**
Related to https://github.com/rapidsai/cudf/issues/1263, we could use support for rolling functions in the unweighted expanding window case - e.g. functions for which each point in the result sequence is dependent on all of the data 'before' it in the source sequence.

https://pandas.pydata.org/docs/reference/api/pandas.DataFrame.expanding.html

- count
- sum
- mean
- median
- var
- std
- min
- max
- corr
- cov
- skew
- kurt
- apply

Since many of these aggregations can be expressed as a recurrence relation they might be computable using the same machinery already being developed for `ewm`.

**Describe the solution you'd like**
Implement expanding windows in cudf/libcudf that produce expanding window scans of the data.

**Describe alternatives you've considered**
Wrap rolling machinery with `window_size=len(data)`, which would likely be inefficient.

**Additional context**
Add any other context, code examples, or references to existing implementations about the feature request here.