Improved Python ergonomics

[leo-ware]

I wrote a couple of helper utilities to make it easier to use the code inside other Python code.

Key changes:

• Added a Dataset.from_pandas method
• Added a train method to IrtModel
• Added an example notebook demonstrating how to use the new features
• Tests for my code

I was trying to use this library for a personal project, but I had trouble understanding how to orchestrate the whole IrtModel/IrtTrainer/IrtConfig setup at first. Also, I needed to use data that was in memory rather than in a file for my usecase, which was tricky previously. So, I thought I'd write some methods to get around these issues.

The following code now works:

from py_irt.dataset import Dataset
from py_irt.models import OneParamLog
import pandas as pd

df = pd.DataFrame({
    'subject_id': ["joe", "sarah", "juan", "julia"],
    'item_1': [0, 1, 1, 1],
    'item_2': [0, 1, 0, 1],
    'item_3': [1, 0, 1, 0],
})
dataset = Dataset.from_pandas(df, subject_column="subject_id", item_columns=["item_1", "item_2", "item_3"])

trainer = OneParamLog.train(data)
print(trainer.irt_model.export())