Fix some wrong synchronization waits in usb_write()

To make the full use of the ping-pong endpoint
Transfer speed before this fix:
616448,618496,618496,615424,615424->616857.6 bytes/s
Transfer speed after this fix:
707584,709632,707584,709632,710656->709017.6 bytes/s (+14.9%)
Tested by running hw status

common_arm/usb_cdc.c

@@ -797,7 +797,7 @@ int usb_write(const uint8_t *data, const size_t len) {
         while (pUdp->UDP_CSR[AT91C_EP_IN] & AT91C_UDP_TXCOMP) {};
 
         UDP_SET_EP_FLAGS(AT91C_EP_IN, AT91C_UDP_TXPKTRDY);
-        while (pUdp->UDP_CSR[AT91C_EP_IN] & AT91C_UDP_TXPKTRDY) {};
+        while (!(pUdp->UDP_CSR[AT91C_EP_IN] & AT91C_UDP_TXPKTRDY)) {};
     }
 
     // Wait for the end of transfer
@@ -812,7 +812,7 @@ int usb_write(const uint8_t *data, const size_t len) {
     if (len % AT91C_EP_IN_SIZE == 0) {
 
         UDP_SET_EP_FLAGS(AT91C_EP_IN, AT91C_UDP_TXPKTRDY);
-        while (!(pUdp->UDP_CSR[AT91C_EP_IN] & AT91C_UDP_TXCOMP)) {};
+        while (!(pUdp->UDP_CSR[AT91C_EP_IN] & AT91C_UDP_TXPKTRDY)) {};
 
         UDP_CLEAR_EP_FLAGS(AT91C_EP_IN, AT91C_UDP_TXCOMP);
         while (pUdp->UDP_CSR[AT91C_EP_IN] & AT91C_UDP_TXCOMP) {};