Remove type parameters (#252)

They're unneeded, and breaking on Julia master:
JuliaLang/julia@feb2988

Author: chriselrod

ext/DistributionsADTrackerExt.jl

@@ -64,12 +64,12 @@ for f = [:hcat, :vcat]
 end
 
 function Base.copy(
-    A::TrackedArray{T, 2, <:LinearAlgebra.Adjoint{T, <:LinearAlgebra.AbstractTriangular{T, <:AbstractMatrix{T}}}},
+    A::TrackedArray{T, 2, <:LinearAlgebra.Adjoint{T, <:LinearAlgebra.AbstractTriangular{T}}},
 ) where {T <: Real}
     return track(copy, A)
 end
 @grad function Base.copy(
-    A::TrackedArray{T, 2, <:LinearAlgebra.Adjoint{T, <:LinearAlgebra.AbstractTriangular{T, <:AbstractMatrix{T}}}},
+    A::TrackedArray{T, 2, <:LinearAlgebra.Adjoint{T, <:LinearAlgebra.AbstractTriangular{T}}},
 ) where {T <: Real}
     return copy(data(A)), ∇ -> (copy(∇),)
 end
@@ -588,4 +588,4 @@ Distributions.InverseWishart(df::Real, S::TrackedMatrix) = TuringInverseWishart(
 Distributions.InverseWishart(df::TrackedReal, S::TrackedMatrix) = TuringInverseWishart(df, S)
 Distributions.InverseWishart(df::TrackedReal, S::PDMats.AbstractPDMat{<:TrackedReal}) = TuringInverseWishart(df, S)
 
-end # module
+end # module