Fix some bugs in H5 writers

@backes noticed some weird bug where the fist element an array we were writing would take the same value as the last element. I traced that down to an off by one error where a > should have been >=.

While I was there I noticed that the output for arrays with rank > 2 seemed a bit bonkers. I fixed that, we probably never noticed because no one is writing higher rank arrays, but we should make sure it does the right thing.

I verified that this doesn't change anything else by dumping a few events to HDF5, and by updating the unit tests so that they would have reproduced the problems here.