re-add workqueue and make WQ or tasklet selectable by module parameter

default to WQ for ARM64
(some SOCs use only one little A53 core for all MSI IRQs and
tasklets on the same core cannot process more than 10 TS streams
in uncached coherent memory)

ddbridge/ddbridge.h

@@ -202,6 +202,7 @@ struct ddb_dma {
 	u32                    div;
 	u32                    bufval;
 
+	struct work_struct     work;
 	struct tasklet_struct  tasklet;
 	spinlock_t             lock; /* DMA lock */
 	wait_queue_head_t      wq;