We need to add Sphinx documentation for PyMongoArrow. Documentation must include:

• API documentation for all public PyMongoArrow APIs
• A 1-page getting started tutorial/guide
• Limitations
• Examples

The examples should cover the following scenarios:

• Dealing with timestamp values in seconds/milliseconds/nanoseconds
• Handling missing data using pd.DataFrame.fillna
• Writing data to Parquet format (e.g. to use with ADL)

The limitations section should:

• List which BSON types are supported and which are not
• Elaborate upon limitations caused by our reliance on PyArrow for supporting conversion to NumPy/Pandas (e.g. ndarrays cannot be created arrays with null values, mixed type columns are not allowed in Pandas.DataFrames even though Pandas supports them)