[FEATURE]: Support more column types for resampling and interpolating

[BrianDeacon]

Is there an existing issue for this?

• I have searched the existing issues

Problem statement

My vanilla use case is wanting to forward fill dates (not the dates used for the time series, but another date value)

Proposed Solution

My read of the code is that "average" aggregation and "linear" interpolation are the only cases that actually require the target columns be numeric. Except for arrays, maps and structs, other data types could be floor'd ceil'd, min'd and max'd for aggregation. Interpolation could ffill and bfill with no problem.

Could interpolate and aggregate be more specific in what column types it forbids based on the type of operation being performed on it? I'd really like to ceil and ffill a date column.

Additional Context

No response

[BrianDeacon]

If the feature seems like it fits with your plans, I could try to take a swing at it with a little advice.

[BrianDeacon]

I've got a draft PR #421

I need to add new tests, but thought I'd check first if the approach or even the feature works for you guys. You'll also see a comment about how backwards-compatible we want the change to be. i.e., do we want to keep "averaging" strings.

Would you prefer I submit the PR and we discuss it within there, or here in the issue?