Best way to add columns to an existing Feather file (>10GB) in terms of performance in python

### Describe the usage question you have. Please include as many useful details as possible.

The current recommendation is: "Now that Feather supports zero-copy reads into Arrow, the preferred approach would be to zero-copy read the file, add a column, then write a new Feather file" from https://issues.apache.org/jira/browse/ARROW-1518.

This approach is quite costly. I'm very interested to know a more performant way. Thank you.

P.S. I assume that I don't need to write usage details, as this problem is quite common if large data is stored in this format and needs to be updated.

### Component(s)

Other

Start by reviewing the current zero-copy read, add-column, and rewrite recommendation cited from ARROW-1518, along with the Python Feather usage described in the issue. Investigate whether a supported, more performant approach exists for files over 10GB; done would be a documented recommendation or a clearly scoped implementation change.

Written by the indexing model from the issue text.