Add "variable" kwarg to load_dataset #59

Allow users to select individual "variables" to load from a given dataset. Some datasets (e.g. RACMO and REMA) combine various files, each containing individual variables. Some may not be required by users and simply increase memory overhead. By adding a variables [list] kwarg to specify variables, we can only load what's needed.

Where possible, extend this across all loaders to only load specified variables (default to all variables), even if loading only a single netCDF file.