Enhancement: Would be great if we can load dspy dataset directly from pandas dataframe

[Anindyadeep]

This would be useful for certain scenarios, where we might not be interested to save the csv (personally this was the case for me) instead directly load from some dataframe.

Here is an example:

dataloader = DataLoader()
train_dataset = dataloader.from_pandas(df=df_train, fields=["question", "query"])