Change the way we test snapshots

[cfcosta]

Uses enzyme-to-json on all places instead of only the ones @jotak changed.

[jotak]

@cfcosta I've already added enzyme-to-json with this PR : #1270 :-(
So I guess you can revert that part of my commit (remove the dependency)

[israel-hdez]

What I don't like too much is that attrributes are not always expanded, like:

<TablePfProvider columns={{...}} striped={true} bordered={true} hover={true} dataTable={true} className=\\"\\" condensed={false} inlineEdit={false} components={{...}}>

note the columns={{...}} and components={{...}} attributes. I also saw the same happening for style attributes.
If enzyme-to-json solves this better, I would prefer to include the dependency.

[cfcosta]

@israel-hdez this happens because it does shallow instead of complete rendering, so it does not saves snapshots when you pass objects/functions to the components.

This is necessary to have some layering on the rendering. We can use full rendering to get those to be saved as well, but that's out of the scope.

[cfcosta]

@jotak @israel-hdez changed the PR to start using enzyme-to-json on all places, can you review it again?

[jotak]

So, this is good to me, but out of curiosity why finally enzyme-to-json was better than just .debug() ?

[cfcosta]

@jotak I'd prefer to keep debug(), but we already have enzyme-to-json on the repo, and it works. The less we change the better.

Also, @israel-hdez found some issues where important data was being hidden. To fix those problems, we would need to use .render().debug(), but it requires redux to be "running" to do that. So I decided to go the simplest way and then we can try something more complex later.