Use lightning.pytorch.loggers.TensorBoardLogger for tensorboard logging.

[Lilferrit]

Instead of using torch.utils.tensorboard.SummaryWriter, it might be better to use PyLightning's tensor board logging interface instead, similar to the implementation in dev_latest_depthcharge. This would mainly (more easily) allow other metrics, such as the optimizer metrics, to be written to the tensorboard logs, but this would also have the benefit of making the implementations of the tensorboard and the soon to be introduced csv logging more consistent with one another.