fix typing of UnivariateZeroState init

[Gyoshi]

eltype(x1)[] is incorrectly used here instead of typeof(x1)[]. Does not create a problem for most cases since eltype(x) == typeof(x) when x is a scalar.

[codecov]

Codecov Report

Merging #208 (a8fe4b0) into master (fb83ad1) will not change coverage.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##           master     #208   +/-   ##
=======================================
  Coverage   88.98%   88.98%           
=======================================
  Files           9        9           
  Lines        1461     1461           
=======================================
  Hits         1300     1300           
  Misses        161      161           

Impacted Files	Coverage Δ
src/derivative_free.jl	89.64% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update fb83ad1...a8fe4b0. Read the comment docs.

[jverzani]

Many thanks! While I have your attention, any thought on how to widen the type signature beyond a union of Tuple and Vector to include any iterable of length 2?

[Gyoshi]

Ok, I am hardly an authority on the matter, but how about not explicitly requiring an iterator, and just using it as if it is one--letting the code throw a MethodError if getindex or iterate is not defined?
JuliaLang/julia#23429 (comment)
If you want to be strict in terms of length, I would just @assert length(arr) == 2, but of course you lose a bit of generality since you are requiring that length has a method defined for that iterator.

If you need to specify that the eltype needs be the same in the type signature (for an arbitrary type of the iterator), you could look into WhereTraits.jl, but I'm not convinced that is a good idea.