Generalise the data representation

[r59q]

There are issues in the architecture regarding how we type-check the data, as a result stuff like magnetic values might become harder, as the generics are currently coded to infer types for the MicrobitAccelerometer data type deeply within the codebase. This should be fixed before adding magnetic sensor data as an option. Maybe having a generic that is known at compile-time isnt the best solution for this. Maybe we should generalize it to a vector of variable length and make implementations that announce their size and description?