AWH friction is wrong in checkpoints for dimensions > 1

Summary

The AWH checkpoint reading/writing of the friction tensor data has always been incorrect for AWH dimensions > 1. For AWH simulations restarting from checkpoints this has resulted in incorrect friction tensors for dimensions > 1 and the corresponding friction metric volume.

This has not affected PMFs, biases or sampling. If calculating a friction, or diffusion coefficient from the friction, it would be apparent that the data is nonsensical for dimensions > 1.

Exact steps to reproduce

Run a two-dimensional AWH simulation and restart from checkpoint. The friction metric and friction tensors will be dramatically different when analysing AWH output before compared to after the checkpoint.

Possible fixes

A fix is on the way.