Get `donatello` ready to run ML model

[JackKelly]

• [x] Install miniconda etc. (or have I already installed it?)
• [x] Install git.
• [x] Git clone power_perceiver
• [x] Copy v15 of the dataset to donatello's SSD

[JackKelly]

Done.

Although it's interesting that the data transfer rate between leonardo and donatello is about a tenth of the speed it could be:

It should manage more like 4 GB/sec. But I'd guess that's just because I was copying using a single thread?