Assumes data matrix is tall or square, but not wide. Does not add intercept.
beta_ols(X, y)
Data matrix (n by p)
n
p
Response vector (length n)
Vector of regression coefficients