Fix bugs in the random seeding of the JER uncertainty

@markowen found that the JER was using the same seed for smearing jets as is done in JetCalibTools. This results in a correlated smearing and a large over-estimation of the uncertainty. Additionally, as most people use MC as pseudo-data it is more correct to use the max(MC,data) resolution. This was not anticipated when the tool was originally written but results in very small changes. Finally advanced users can set the seed themselves, but this currently set the seed the same for all jets which is not a usual use-case so this has also been modified.