setting generation-seed during injections in gaussian noise leads to baised pp-test

Setting generation-seed when trying to run PE on injections in gaussian noise results in the injections being placed in the same data. This can lead to biased pp-plots (see attached images below). This may not be an issue, but maybe we can log a warning.

@colm.talbot 's thoughts

I would rather that a single seed can be provided and it will generate different data for each event.