The org.openjdk.bench.java.net.URLEncodeDecode microbenchmark accidentally generates strings with a lot of \u0000 chars, leading to skewing the results of encoding in particular: all tested strings need a lot of encoding, mainly replacing \u0000 with %00

This can be improved. At the same time it would be good if the benchmark had some parameters to test different proportions of inputs that needs to be encoded or not, so that we avoid designing optimizations that overfit for one or the other.