anagram: Make the test suite as flexible as possible

[rbasso]

For a while, we are doings things that seem to increase the number of solutions accepted by the test suites:

• In Add -fno-warn-type-defaults to supress unwanted warnings. #144 we fixed some exercises to allow polymorphic solutions to compile without warnings.
• In tests: Tracking migration to hspec with fail-fast. #211, we rewrote most of the test suites. In the process, some of the tests were changed to allow more general numeric types.
• In crypto-square: Completely rewrite tests. #391, we changed crypto-square to stop testing for intermediary functions, and test only for the essential properties of the solution.
• In anagram: Change tests to be order-insensitive #392, we recently agreed that we should ignore the order of the words returned by anagramsFor, because the order is not an essential part of the problem.

All of those things where based on the same idea: To allow more diverse solutions.

This seems to be aligned to what was discussed in exercism/discussions#41 and exercism/discussions#44.

Taking this process to the next level, if the essence of the anagram problem is to find anagrams, there is no reason to restrict the solutions to work only on Strings and Lists. The following signatures seem equally reasonable:

anagramsFor :: String ->          [String] ->          [String]
anagramsFor :: String -> Sequence  String  -> Sequence  String
anagramsFor :: String -> Set       String  -> Set       String
anagramsFor :: Text   ->          [Text]   ->          [Text]
anagramsFor :: Text   -> Sequence  Text    -> Sequence  Text
anagramsFor :: Text   -> Set       Text    -> Set       Text

This PR is an experiment on trying to write the most general possible test suite, allowing users to create beautiful solutions using important libraries that are usually not allowed in most of the exercises.

Closes #392.

[petertseng]

note for my study - this allows either Text or String to be used. Interesting.

[rbasso]

IsList is fantastic. Way more flexible than I imagined. 😄

[petertseng]

Seems good! I hope the many fromList and toList do not make it too hard for students to understand what is going on. That will depend on having their feedback so it seems like it is time to merge this.

[petertseng]

I manually verified that this works on also String -> Set String -> Set String and String -> [String] -> [String]. I kind of wish that validation could be performed automatically. For example, if we would have multiple example files, and our travis.yml would test each one of them in turn.

[rbasso]

Hummm...this is a great idea, @petertseng.

I was thinking in something similar to solve the problem of having a separated package.yaml for the example. I think that maybe having multiple examples would automatically solve that.

I'll try to think in a way to test multiple solutions on Travis.