[ENH] Add Fama-French 3-factor and 5-factor expected return models

[WilliamTDavies]

Closes #595
Closes #634

Summary

Adds Fama-French three-factor and five-factor expected-return estimation.

Implementation

• Adds ff_return() with support for model="ff3" and model="ff5".
• Estimates asset factor loadings using OLS and historical mean factor premia.
• Uses the regression intercept when estimating loadings but does not treat it as expected abnormal return.
• Aligns asset and factor observations by date and handles missing observations independently for each asset.
• Requires simple factor returns expressed as decimals and rejects incompatible log returns.
• Adds ff3_return and ff5_return dispatch through return_model().

Documentation

• Documents the FF3 and FF5 specifications.
• Documents required factor columns and input units.
• Updates the expected-returns feature list.

Tests

Tests cover:

• FF3 and FF5 estimation.
• Recovery of known synthetic factor loadings.
• return_model() dispatch for both models.
• Compounding and non-compounding annualisation.
• Missing required columns and invalid model names.
• Non-overlapping and missing observations.
• Independent handling of missing asset data.
• Rejection of log returns.
• Additional unused factor columns.

Local validation: 35 expected-return tests pass, and all relevant pre-commit hooks pass.