mavlink main return main loop delay proper size

[dagar]

Going back to @pavel-kirienko's NuttX profiling one thing he found was > 1% of total system time was spent on this line in mavlink. #6829 (comment)

This turned out to be get_main_loop_delay() returning a 64 bit integer, although the loop delay itself is an unsigned int. Simply fixing the return type to match the data reduces cpu by 1-2%.

Before

 185 mavlink_if0                 19839 10.658  1736/ 2380 100 (100)  w:sig 
 186 mavlink_rcv_if0               763  0.429  1608/ 2140 175 (175)  w:sem 
 190 mavlink_if1                  4878  2.646  1656/ 2420 100 (100)  READY 
 191 mavlink_rcv_if1               757  0.429  1320/ 2140 175 (175)  w:sem 
 206 mavlink_if2                 13381  7.081  1720/ 2364 100 (100)  w:sig 
 207 mavlink_rcv_if2               769  0.357  1504/ 2140 175 (175)  w:sem 
 247 mavlink_if3                 12008  6.151  1648/ 2388 100 (100)  READY 
 248 mavlink_rcv_if3               777  0.429  1320/ 2140 175 (175)  w:sem 

After

 185 mavlink_if0                 10836  9.942  1736/ 2380 100 (100)  READY 
 186 mavlink_rcv_if0               449  0.432  1496/ 2140 175 (175)  w:sem 
 190 mavlink_if1                  2623  2.377  1656/ 2420 100 (100)  w:sig 
 191 mavlink_rcv_if1               443  0.432  1320/ 2140 175 (175)  w:sem 
 206 mavlink_if2                  7237  6.556  1720/ 2364 100 (100)  w:sig 
 207 mavlink_rcv_if2               451  0.432  1320/ 2140 175 (175)  w:sem 
 247 mavlink_if3                  6264  5.475  1664/ 2388 100 (100)  w:sig 
 248 mavlink_rcv_if3               447  0.432  1464/ 2140 175 (175)  w:sem 

[TSC21]

How simple things can just ruin performance! Nice one @dagar