Document building needs a rethink.

[NathanCummings]

Currently, Jupyter book will run all of the cells when you try to build the docs. This means that, for example, the cell that shows getting and installing s5cmd, will actually run those commands, downloading the s5cmd source and try to run it to download data.

[NathanCummings]

We are also thinking that we shouldn't be building the docs and tracking the build artefacts with git, rather we should be dynamically building the docs in CI and just tracking the Jupyter notebooks. This would reduce the size of the repo significantly and make it quicker to clone.

This may have implications in that we'd possibly need to build the docs locally if doing development work that doesn't involve the docs, which is not entirely ideal. To be investigated.