Support the bit-shuffling changes from llama.cpp

[philpax]

A new file version is being introduced to change how the tensors are stored on-disk: ggerganov/llama.cpp#1305

We will need to support this version, as well as the older versions.

[philpax]

It's been merged: ggerganov/llama.cpp#1405

There doesn't seem to be a migration path at present, so let's wait a bit: ggerganov/llama.cpp#1408

[philpax]

This is done in #226, but I'd like to set up a migration path before I close this

[philpax]

No migration path for now. See #261