Fix type stability of Statistics.var with dims.

[maleadt]

Fixes #336

[codecov]

Codecov Report

Merging #337 into master will increase coverage by 0.01%.
The diff coverage is 50.00%.

@@            Coverage Diff             @@
##           master     #337      +/-   ##
==========================================
+ Coverage   79.44%   79.46%   +0.01%     
==========================================
  Files         156      156              
  Lines        8908     8910       +2     
==========================================
+ Hits         7077     7080       +3     
+ Misses       1831     1830       -1     

Impacted Files	Coverage Δ
test/statistics.jl	0.00% <0.00%> (ø)
src/statistics.jl	88.88% <100.00%> (-11.12%)	⬇️
src/pool.jl	79.04% <0.00%> (+1.42%)	⬆️

---

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 2fbf594...ae1283d. Read the comment docs.

[jumerckx]

With this, var and std don't work with tuple-typed dims.
e.g.

var(cu(rand(10, 10, 10)), dims=(2, 3))

ERROR: MethodError: no method matching getindex(::Tuple{Int64,Int64,Int64}, ::Tuple{Int64,Int64})
Closest candidates are:
  getindex(::Tuple, ::Int64) at tuple.jl:24
  getindex(::Tuple, ::Real) at tuple.jl:25
  getindex(::Tuple, ::AbstractUnitRange{#s69} where #s69<:Real) at range.jl:290
  ...
Stacktrace:
 [1] _var(::CuArray{Float32,3}, ::Bool, ::Nothing, ::Tuple{Int64,Int64}) at c:\Users\jules\Documents\code\Transformer\Untitled.jl:8
 [2] var(::CuArray{Float32,3}; corrected::Bool, mean::Nothing, dims::Tuple{Int64,Int64}) at D:\buildbot\worker\package_win64\build\usr\share\julia\stdlib\v1.4\Statistics\src\Statistics.jl:335
 [3] top-level scope at REPL[1]:1

[maleadt]

OK, should be fixed in #393