Probably just leftovers from a long time ago.
Anode to MCU, cathode to ground is a more intuitive configuration, since you have to set the pin high to enable the LED. This is probably why this was chosen for the user LED.
The other way around is technically slightly better, since MCU outputs are much better at sinking the current than sourcing it. This difference is negligible in CMOS logic, but was significant in TTL. So people that remember those days still do "anode to VDD" configuration.
The USB IC is not user accessible, so it does not really matter which way LEDs are connected.