Context. A wide sensor table — you only care about certain columns and want them in a specific order.
Your task. Write pick_columns(matrix, cols) that takes a 2D list and a list of column indices, and returns a NumPy array with just those columns, in the order given.
Example. matrix=[[1,2,3],[4,5,6]], cols=[0,2] → array([[1, 3], [4, 6]]).
Notes. cols=[2,0] returns columns in that swapped order, not sorted. cols=[] returns a 2D array with 0 columns (one empty row per input row).