feat(api): make topk() and value_counts() more flexible

[NickCrews]

This PR changes many things. I bet we don't want all of them, but I thought it would be easiest to just put up a laundry list and then I will prune out the things we don't like.

• Add a Table.topk(). I want this frequently to check for duplicates. I don't think it makes sense to provide a by argument here?
• Add "See Also" links to the docstrings
• made the "k" param to topk() be optional. If you don't supply one, it just ranks the values in descending order without limiting them. I do this often in interactive analysis, I like less typing, the .head(10) during the repr in interactive mode takes care of it for me. This isn't breaking for anyone.
• Makes the default column name for Column.topk() be {column_name}_count, which is the same as for Column.value_counts(). This default is better because it is more consistent, and because it allows you to use .column syntax in subsequent expressions, eg col.topk().col_n.max(), where the current default with the parenthesis makes this impossible. For more complex by clauses, this suffix doesn't make sense, so I left the current behavior as is. This IS a breaking change for people relying on the old generation scheme, but IDK, they probably shouldn't have been relying on it. We could in fact add a note saying "consider this unstable" going forward, so we are more free later to change this again.
• improved the top-line docstring for Column.topk(). The old Return a "top k" expression. is self-referential and fairly useless.

With all these changes, Column and Table both have .topk() and .value_counts(), and they all behave consistently, except Column.topk() has a by param, and Table.topk() does not.

[cpcloud]

+1 to all the changes here. I'm guessing that someone may eventually ask for a custom by with the table topk (not entirely sure for what at the moment though), but we can leave it as just count for now.

Thanks for working through this!