Fix #91 by removing the shortcuts from the fitting predicates

[sjakobi]

This version passes the fusion property tests using the Unbounded layout.

[sjakobi]

What I don't understand is why it doesn't seem to be necessary to continue checking for the absence of SFail when using the Unbounded layout – at least the property tests don't seem to mind (and neither does dhall!)

[sjakobi]

The advantage of this solution is that no types need to change. However IMO FittingPredicates are overkill for the Unbounded layout, and I suspect that performance for bounded layouts will be affected too.

IMO having a separate layoutUnbounded function as an alternative to layoutSmart and layoutPretty would still be the more principled approach (see #92 (comment)).

With that approach, FittingPredicate would

• …lose the Maybe on the maxWidth paramater
• in layoutWadlerLeijen the FittingPredicate parameter would be replaced with a selectNicer function, that would be very simple in the case of layoutUnbounded, and could be constructed from layout{Pretty,Smart}'s FittingPredicates
• Unbounded should be removed as a PageWidth constructor, but could stay if we want to accept some internal clunkiness for compatibility reasons

IMHO Unbounded should really die though: It negates the essence of what layoutPretty and layoutSmart are.

[sjakobi]

• Unbounded should be removed as a PageWidth constructor, but could stay if we want to accept some internal clunkiness for compatibility reasons

Ah, I had missed some uses of PageWidth, for example in the WithPageWidth Doc constructor, where layoutCompact relies on Unbounded…

[sjakobi]

Closing in favour of #98.