Standardize set formatting

[jamescurtin]

Creating this PR as a suggestion: please feel free to disagree & close! 🙂

Previously, sets with fewer than 5 keys were represented on a single line, whereas larger sets were broken up with one element per line. Now, all sets are serialized with the latter behavior.

I see two cases where always expanding sets is useful:

1. When the key elements are long & the length of the line exceeds the recommended ~80-90 columns, despite there being fewer than 5 elements
2. When a set has fewer than 5 elements, but more are added. Having consistent formatting rules will result in marginally smaller diffs.

[joshtemple]

Thanks for the suggestion! I'd like to keep this feature in place because it results in more human-looking LookML and decreases file length.

For line width, I would be interested in adding a feature to the serializer that adds line breaks when needed to constrain the width of the output LookML, which would solve that issue. I created #23 to track this. Would be interested in hearing your thoughts over there.

I hear what you're saying about diffs, but I think the amount of scrolling it takes to get through a LookML file is more important to me from a UX standpoint than the length of the diff in the less likely case where a set goes from 5 > 5+ elements.

Also, I realized from commit 7694406 that I need to update the name of that second test! It's not actually a duplicate, just has a duplicate name (it tests behavior with a name key added).

[jamescurtin]

👍 Thanks for taking a look!