save aligned data for faster data loading

[krsnapaudel]

1. Align with crop calendar
2. Align indices of labels and predictors
   I realized that aligning with lead time must be done afterwards. So the current code in trim_to_lead_time must be split into two parts: do 1 and 2 above and save data. Do the trimming to lead time later.

[krsnapaudel]

The code to save and load aligned data was removed in a subsequent merge. Reopening.

Considerations:

1. Data path vs aligned data. Is the aligned data path inside data path?
2. Should we reconsider how data is organized now, i.e. crop and then country?
3. We should also save data after TorchDataset interpolates and aggregates time series. Should we make it a separate issue?

I have second thoughts about saving aligned data because it would be hard to separate aligning to crop season and trimming time series based on lead time.