Document nil string storage anti-pattern

Nobody has claimed this yet.

This kind of question has come up quite frequently:
https://gitter.im/luckyframework/Lobby?at=5f4fc74f36e6f709fd1ca322

It boils down to storing "empty" to represent a nil/falsy value as being inadvisable, both from a database perspective (querying in the default PSQL CLI doesn't differentiate between the display of an empty string and a NULL value) and from a Crystal perspective (you can't do nifty things like what Paul noted in the Gitter chat above).

In case something happens to Gitter, here's the essence of what we want to call out (from Paul):

> Some data are required, like note but note can be "".
This doesn't mean it's empty, because it's a string, not a nil
▸ note: is required
note is on permit_columns

> I'd recommend keeping it and instead treating an empty string as nil. Otherwise you may do stuff like: span @user.name and it'll print an empty string. It's also easier to check for nil than an empty string because you can use try. @user.name.try { |name| span name } for example

We should add a callout in the documentation on the website around this being an anti-pattern from the Lucky perspective, the benefits of avoiding it, and how to work around it (`my_column.errors.reset_errors` in `before_save`) if desired.

Locate the relevant documentation section on the Lucky website and review how database values and nil handling are currently explained. Add a callout covering the empty-string anti-pattern, its database and Crystal drawbacks, and the documented my_column.errors.reset_errors workaround; done means the guidance is clear and published in the appropriate section.

Written by the indexing model from the issue text.