Add non-splatting variants that speed range tests with inline=no

When testing coverage, we set --inline=no, and the test/ranges.jl file
often times out. With this change, the tests are still quite slow,
but seem unlikely to time out. (On a reduced version of the bottleneck
in those tests, the speed improvement was nearly 20x.)

base/abstractarray.jl

@@ -734,10 +734,18 @@ cat{T}(catdims, A::AbstractArray{T}...) = cat_t(catdims, T, A...)
 
 cat(catdims, A::AbstractArray...) = cat_t(catdims, promote_eltype(A...), A...)
 
+vcat(A::AbstractArray) = cat(1, A)
+vcat(A::AbstractArray, B::AbstractArray) = cat(1, A, B)
 vcat(A::AbstractArray...) = cat(1, A...)
+hcat(A::AbstractArray) = cat(2, A)
+hcat(A::AbstractArray, B::AbstractArray) = cat(2, A, B)
 hcat(A::AbstractArray...) = cat(2, A...)
 
+typed_vcat(T::Type, A::AbstractArray) = cat_t(1, T, A)
+typed_vcat(T::Type, A::AbstractArray, B::AbstractArray) = cat_t(1, T, A, B)
 typed_vcat(T::Type, A::AbstractArray...) = cat_t(1, T, A...)
+typed_hcat(T::Type, A::AbstractArray) = cat_t(2, T, A)
+typed_hcat(T::Type, A::AbstractArray, B::AbstractArray) = cat_t(2, T, A, B)
 typed_hcat(T::Type, A::AbstractArray...) = cat_t(2, T, A...)
 
 # 2d horizontal and vertical concatenation

base/base.jl

@@ -35,6 +35,7 @@ call(T::Type{GenSym}, n::Int) = Core.call(T, n)
 call(T::Type{WeakRef}) = Core.call(T)
 call(T::Type{WeakRef}, v::ANY) = Core.call(T, v)
 
+call{T}(::Type{T}, arg) = convert(T, arg)::T
 call{T}(::Type{T}, args...) = convert(T, args...)::T
 
 convert{T}(::Type{T}, x::T) = x