v0.4.0

• Polars backend returns WRDS date columns as Date (#66): WRDS calendar-date columns (datadate, trd_exctn_dt, CCM link dates, FISD dates) are now cast to polars.Date instead of polars.Datetime, matching the R package.
• Added FRED-MD and FRED-QD macroeconomic databases: download_data("FRED", "FRED-MD" / "FRED-QD") download the McCracken and Ng (2016, 2021) macro panels. transform=True applies each series' transform code; vintage enables point-in-time analysis.
• Added Global Factor Data, Pastor-Stambaugh, and Stambaugh-Yuan downloads: download_data("Global Factor Data") downloads portfolios, industries, or cutoffs from Jensen, Kelly, and Pedersen (2023). download_data("Pastor-Stambaugh") and download_data("Stambaugh-Yuan") download the liquidity and mispricing factors.
• OSAP download aligned with beginning-of-month and scaled returns: download_data("Open Source Asset Pricing") now uses beginning-of-month dates (was end-of-month) and decimal returns (divided by 100).
• sorting_variable is now optional for factor_library: it returns the default construction for all sorting variables when omitted, and passing None for a filter column removes that filter.
• Added detail parameter to estimate_fama_macbeth: detail=True returns coefficients plus summary statistics (mean r_squared, adj_r_squared, n_obs). Default unchanged.
• Dependencies (replaced pyfixest with formulaic): dropped pyfixest for formulaic.