Tommey there are a couple of different ways that you can go.  The customary way to build large Li ion packs is to place a bunch of cells in series, and then place those series stacks in parallel.  There are many charge controller products out there that will manage serial stacks, typically up to six cells per chip.  They vary by how they handle the charge balancing.  Serial stacks depend on all the cells in the stack performing pretty uniformly.  Since you are getting used cells the serial stack arrangement could be troublesome. 

In order to use cells that are disparate you pretty much need a charge controller per chip.  How you manage that comes down to efficiency and economy.  Linear charge controllers are compact and cheap:  ~$.50/ cell but are wasteful and give off considerable heat.  They need typically at least 4.75V and burn up whatever does not go into the cell.  So when your cells are down at 3V, the controllers will dissipate 65% as much power as you deliver to the cell.  There are switching controllers, but the price jumps up to about $1.5 / cell plus the additional resistors and capacitors and the inductor.

On the matrix idea, I would do a single high side switch per cell based on a pair of back to back MOSFETs.  That way you can block both charge and discharge.  If you don't want to build an auxiliary bias supply you will need to use P channel devices controlled by a logic output that can swing 4.2V or a small open drain N channel device.  This requires 16 I/Os for 16 cells, but that is easily managed with an I2C expander.  There are some pretty good P channel devices out there that are fairly low cost. 30 mOhm parts run about $0.22 each in hundred qty.  That's OK to about 2A.  If you want to discharge at 5A or 10A then you will need better transistors.