Introduce Univ newtype

[phadej]

This is first step for https://github.com/dmwit/universe/issues/23

I introduce Univ newtype, so instance definitions could be done based on combinators on Univ.

Also I change indentation to use spaces.

Follow-ups would be:

• Don't use universe or universeF in instance declarations
  • Introduce needed combinators to Data.Universe.Helpers
• Experiment on Univ implementation

Open questions:

• [x] universe and universeF don't need to be in the class anymore. Then we won't accidentally memoise when using them!
• [x] Monoid instance for Univ (using fair interleaving?)
• [x] Foldable/Traversable for Univ (Foldable provides natural toList!)
• [ ] better name for Univ, maybe Distinct or USet?
• [ ] Rename Data.Universe.Helpers to Data.Universe.Univ

Some answers:

• I bumped versions to 2 as this is huge update.
• I think that small breaking changes in universe-base (i.e. combinator removal, or even redefining of Univ) won't break universe-instance-* API. As Univ isn't exported from instance packages, their public api isn't broken (if they continue to work). So we can depend on universe-instances-* >= 2 && <3. That could be documented in universe package haddock if this scheme makes sense.

[phadej]

@dmwit, I also commited .travis.yml into repository, with cabal-install-1.24 it's possible to build multi-package projects using "standard tools". .travis.yml has rough edges, new-build isn't yet polished, but having CI would be great!

[phadej]

Note Univ is different implementation of Omega in https://hackage.haskell.org/package/control-monad-omega-0.3.1/docs/Control-Monad-Omega.html

By making similar Applicative instance, we can have similar definitions for Universe and Finite of tuples for example.