Add support for --ignore-on-update

[jkeuhlen]

I think the README update is a decent description of what I'm trying to achieve here:

You can also use the `--ignore-on-update` flag to mark columns that should be ignored during subsequent imports (useful for timestamps that shouldn't change):

# Export all data but ignore created_at and updated_at columns during updates
ripoff-export --ignore-on-update created_at --ignore-on-update updated_at /path/to/export

This adds a `~ignore_on_update` metadata field to rows containing the specified columns. During import, these columns will be included for new rows but ignored when updating existing rows, preventing timestamp-only changes from triggering unnecessary updates.

Generally, when using ripoff I tend to do the following:

• Work on a feature in my normal way.
• When a feature is finished, I want to persist the locally created DB state as my known "good" state for future development.
• run ripoff-export and do any manual tweaks that are needed.
• In the future, I can run ripoff db/fake-data to reset to my known good state for development.

One annoying thing (that this PR attempts to solve) is that I use created_at/updated_at timestamps on every table in my database so each ripoff-export invocation touches every table definition in my saved dataset.

To keep the diffs more minimal, --ignore-on-update provides a way to skip specifically named columns (that, in this case repeat across tables).

[jkeuhlen]

Just ignoring the other executable this builds.

[jkeuhlen]

Going to take another pass at this from a different direction (excluding certain fields on export entirely)