Add rolling min, max, median, and quantile methods

Normalized profile

What this task asks for

Add the remaining rolling window methods to Expr and Series with consistent validation and backend delegation. When `min_samples` is `None`, it defaults to `window_size`. For lazy backends, this operation requires `.over(order_by=...)`. `quantile: float` -- The quantile to compute, must be in [0, 1]. Out-of-range values raise `ValueError` with message starting with `"Quantile must be between 0.0 and 1.0"`. `interpolation: str` -- Interpolation method when the quantile lies between two data points. One of: `'linear'`, `'lower'`, `'higher'`, `'nearest'`, `'midpoint'`. Invalid values raise `ValueError` with message starting with `"Interpolation must be one of"`. DuckDB does not support `percentile_cont` as a windowed aggregate function; rolling_quantile with `.over()` is not available on DuckDB. For lazy backends (Polars, DuckDB, Dask), rolling operations must be followed by `.over()` with `order_by` specified.