Add S3 support for dataset reading and saving

**Is your feature request related to a problem? Please describe.**
We need to support reading and saving data to S3 rather than only disk support

**Describe the solution you'd like**
When starting up Tanuki, it should be configurable where data is saved or read from. That should be used to save align statements and patch datapoints and read data for finetuning apart from using local disk