Volatility forecasting and VaR backtests

Toolset vol_forecast: 3 tools.

Tool	What it does	Notes
`vol_rolling_forecast`	Backtests a volatility model out of sample: re-estimates it on a rolling or expanding window (every refit_every steps), forecasts the conditional mean, volatility, Value-at-Risk and expected shortfall horizon steps ahead, and scores them against the realized returns.
`vol_forecast_origins`	Forecasts a fitted volatility model from every origin since start, each h steps ahead, with the parameters estimated once (in-sample forecasts, the hedgehog view of how forecasts revert to the long-run level).
`vol_scenario_forecast`	Simulates a fitted volatility model forward from today under a scenario: shocks bootstrapped from the standardized residuals of a chosen historical window (crisis replay) or drawn from the fitted distribution and scaled.

Inputs, limits and outputs are described in Fincept Volatility. Full schemas: fincept_describe_tool.