handle wrap-arouns in time procesing

[divanchykhin]

Currently we admit that time-representing value can wrap-around with time. In this case, if certain time period is relatively small (less than 1/2 of time counter cycle), time point represented as 0xF0 is rather before than that represented as 0x01; respectively, simple comparison 0x01 < 0xF0 will rather be misleading. One way to address this problem is to (1) reserve value 0xFF to denote "infinite future"; (2) assume that measured intervals are less than 1/2 of time counter period; and (3) compare values based on their difference in GF(time counter period) rather than based on their correspondence as integers