On p4, the clock was configured in the wrong order causing a dropped
count on the first transmission. And gdma eof event fails to trigger
parlio_tx eof event when transmitting single byte.
There's two GDMA groups on ESP32P4, one is connected to AHB bus,
and another one is connected AXI bus.
We now have two seperate APIs for allocating DMA channels,
depends on the bus type.
