Fix splatting

[alyst]

With my model I have hit the limit of the arguments that splatting supports (in particular, I was getting StackOverflowError in matrix derivative function).
So I've fixed the problematic spatting instances (op(data...)) with the faster alternative reduce(op, data).
In particular, it is much faster for special cases like hcat, because there is a specialized version of reduce(hcat, x) .

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 67.28%. Comparing base (51ca4a9) to head (d0b2fee).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #167   +/-   ##
=======================================
  Coverage   67.28%   67.28%           
=======================================
  Files          51       51           
  Lines        2494     2494           
=======================================
  Hits         1678     1678           
  Misses        816      816           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[Maximilian-Stefan-Ernst]

Thanks a lot for these fixes. I have one small suggestion - when I benchmark the code, the mapreduce version takes quite long for some reason, so I swapped the order (transpose after concatenation).

using BenchmarkTools

a = [rand(10) for _ in 1:100]

transpose(reduce(hcat, a)) == mapreduce(transpose, vcat, a) # true

@benchmark transpose(reduce(hcat, $a))  # 923.459 ns

@benchmark mapreduce(transpose, vcat, $a) # 49.851 μs