fsi: Fix a race condition resulting in some FSI accesses being lost.

Previously, back to back FSI transactions could happen before teh FSI state machine was complete.
This ensures that the FSI master does not ahve an active transaction before starting a new one.

This could result in logs such as:
access_fsi_mem(): address 0x002820, data: 0x00000000 sta: 0x41000001
access_fsi_mem(): address 0x002804, data: 0x00000000 sta: 0x12000000
vs
access_fsi_mem(): address 0x002820, data: 0x00000000 sta: 0x41000001
access_fsi_mem(): address 0x002804, data: 0x04c04000 sta: 0x41000001

As a result, the CPU often failed to start.

fsi.c

@@ -61,6 +61,11 @@ static int access_fsi_mem(uint8_t slave_id, uint32_t address, fsi_data_length_t
         return -1;
     }
 
+    // Wait for any running operation(s) to complete
+    while ((read_openfsi_register(OPENFSIMASTER_BASE, FSI_MASTER_REG_STATUS) & (FSI_MASTER_CTL_CYCLE_START_MASK << FSI_MASTER_CTL_CYCLE_START_SHIFT)))
+    {
+    }
+
     // Set up request
     word = read_openfsi_register(OPENFSIMASTER_BASE, FSI_MASTER_REG_SID_ADR);
     word &= ~(FSI_MASTER_SID_SLAVE_ID_MASK << FSI_MASTER_SID_SLAVE_ID_SHIFT);