[FEA] Consolidate shared method between DataFrame and Series into Frame

[isVoid]

DataFrame and Series shares much similarity in structure. Operations that works on DataFrame can work (without much adaptation) in Series context, thus, we should consolidate these methods into Frame to reduce duplication.

Caveats:

• Frame base class is also inherited by Index class. This consolidation would allow extra methods that can be operated on Index. It should also be pointed out that operation that depends on the frame having an index cannot be consolidated (such as sort_index).
• After consolidation, the method may have shared docstring for both DataFrame and Series. As pointed out by @vyasr in Refactor Frame scans #9021 (comment). Some Pandas API docs also does this by including examples for both dataframe and series in the same docstring section. It's generally more preferable to have less code duplication to have longer docstrings.

Potential candidates:

• __array_ufunc__
• _concat (DataFrame is by necessity substantially more complex due to the need to handle multiple columns, not all of which may exist in all inputs)
• _n_largest_or_smallest
• add
• append
• argsort
• astype
• describe (The _describe_categorical helper function needs access to value_counts functionality. Unless we want to port something similar to the column layer (and this would be difficult because it involves a groupby), we need the objects passed to this function (and therefore all other _describe_* helpers) to be Series, not Column, which means that the describe needs to remain implemented on Series and have DataFrame call the Series method per column).
• drop
• explode
• floordiv
• from_pandas (The requirements of pulling data out of a pandas Series vs a pandas DataFrame are different enough that we can't consolidate that much common code here)
• groupby
• iloc
• index
• isin (Decided against consolidation, the behavior of DataFrame is too different and Series/Index are one-liners)
• iteritems
• loc
• memory_usage
• mod
• mode (Requires value_counts, same issue as describe)
• mul
• nlargest (merged to _n_largest_or_smallest)
• nsmallest (merged to _n_largest_or_smallest)
• pow
• quantile (The Series method is quite simple and already uses _from_data, and most of the logic revolves around handling the return type so there isn't much overhead added when calling it from the DataFrame API. Attempting to combine them ends up resulting in longer code, especially since their docstrings have slightly different parameters).
• radd
• reset_index
• rfloordiv
• rmod
• rmul
• rpow
• rsub
• rtruediv
• sort_values
• sub
• to_dict
• to_pandas (This function relies on corresponding column methods that return Series objects. Since pandas is not based on a columnar store but rather a block manager, there is no analogous approach to initialize different type of pandas "Frame"s from a single type of data)
• truediv
• values

[vyasr]

As of #9021 I moved forward with the pandas approach of including both Series and DataFrame examples in the same docstring (in Frame), so we can standardize on that approach for the methods mentioned in this issue.

[github-actions]

This issue has been labeled inactive-30d due to no recent activity in the past 30 days. Please close this issue if no further response or action is needed. Otherwise, please respond with a comment indicating any updates or changes to the original issue and/or confirm this issue still needs to be addressed. This issue will be labeled inactive-90d if there is no activity in the next 60 days.