RFC: fix inconsistent thunk API on unique / flatten / flattenDeep

[roxdavirox]

Background

Three functions in the array module require an unnecessary extra call before receiving data:

unique<number>()([1, 2, 2, 3])   // extra ()
flatten<number>()([[1], [2]])    // extra ()
flattenDeep<number>()([1, [2]]) // extra ()

Every other function in the module follows the curried data-last style without a leading thunk:

map(fn)(arr)
filter(predicate)(arr)
groupBy(fn)(arr)

This is tracked in issue #1.

Proposed change

Remove the thunk and accept the array directly:

unique([1, 2, 2, 3])       // [1, 2, 3]
flatten([[1, 2], [3, 4]])  // [1, 2, 3, 4]
flattenDeep([1, [2, [3]]]) // [1, 2, 3]

This is a breaking change (minor in terms of impact — the thunk was useless, callers just need to remove one ()).

Questions for discussion

1. Should this be released as a patch (the thunk was a bug), or does it deserve a minor/major bump?
2. Are there legitimate use cases for the thunk form that we are not seeing?
3. Should the release include a codemod or migration note?

Related: #1