fix extrema

[SimonDanisch]

I could have sworn that I already fixed this.
I'm not sure why this only turns up in a couple of cases - I can't even construct a test case for this.
But for some Arrays, mapreduce isn't able to correctly take the first element of the array as the reduce variable.
I remember debugging this already and it lead me to quite a bit of dark magic in the code of mapreduce.

[codecov-io]

Codecov Report

Merging #338 into master will increase coverage by <.01%.
The diff coverage is 83.33%.

@@            Coverage Diff             @@
##           master     #338      +/-   ##
==========================================
+ Coverage   93.36%   93.36%   +<.01%     
==========================================
  Files          37       37              
  Lines        2711     2712       +1     
==========================================
+ Hits         2531     2532       +1     
  Misses        180      180

Impacted Files	Coverage Δ
src/FixedSizeArrays.jl	41.66% <83.33%> (+0.98%)	⬆️

---

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 11ec063...ae51586. Read the comment docs.

[coveralls]

Coverage increased (+0.002%) to 93.363% when pulling ae51586 on sd/extrema into 11ec063 on master.