Partial inputs - Pakistan flood tutorial

[lillythomas]

This is a tutorial following the style of #149 to show how to generate embeddings from partial inputs and find a signal for a major monsoon flood in Padidan, Pakistan (August 2022) with PCA and t-SNE.

Modifications (aside from AOI and time):

• Use of SWIR band group
• Use of the correct mean and std for the nir, not nir08 band
• Addition of t-SNE computation
• modification to the datamodule that enables the non-normalized dates to be added to the batch dictionary (for plotting with the embeddings when there are 2 or more components involved...aka, date can't be one of the axes)

[weiji14]

We are already returning the non-normalized date as a YYYY-MM-DD string no? Instead of using ts.append(batch["timestep_non_norm"]) in your partial-inputs-flood-tutorial.ipynb notebook, could you use ts.append(batch["date"])?

[lillythomas]

Fixed this. It worked, thanks!

[weiji14]

We should be shuffling the mini-batches in the train_dataloader, is there a reason to set it to False here?

Suggested change

	shuffle=False,
	shuffle=True,

[lillythomas]

I rolled back changes to the datamodule in this PR, so it is the same as in main. I think this should be examined in a separate PR since no training is occurring for the purpose of this one.