PCA: projection

Implement:

def pca_fit(input: np.ndarray, n_components: int) -> dict: ...
def pca_transform(model: dict, input: np.ndarray) -> np.ndarray: ...

• pca_transform centers input using the model's stored mean, not input's own mean, pca_transform is meant to work on new data too, which shouldn't shift the projection's coordinate system.