Use parallel testing infrastructure from CUDA.jl (#2619)

Author: vchuravy

Project.toml

@@ -47,7 +47,7 @@ DynamicPPL = "0.35, 0.36, 0.37"
 EnzymeCore = "0.8.13"
 Enzyme_jll = "0.0.201"
 GPUArraysCore = "0.1.6, 0.2"
-GPUCompiler = "1.6"
+GPUCompiler = "1.6.2"
 LLVM = "9.1"
 LogExpFunctions = "0.3"
 ObjectFile = "0.4, 0.5"