steady_clock should be used for measuring intervals

[danielhochman]

steady_clock cannot go backwards. the system_clock can move with NTP, leap second, etc. durations observed by comparing timestamps from the system_clock will be incorrect if they span a system clock change.

in reality the library will need to use both clocks. start_timestamp should come from the system_clock and duration_micros should be measured using steady_clock.

lightstep-tracer-cpp/src/c++11/lightstep/util.h

Line 11 in e74ebcf

typedef std::chrono::system_clock Clock;

[jmacd]

Acknowledged. I'll follow up.

[jmacd]

Fixed & released:
https://github.com/lightstep/lightstep-tracer-cpp/releases/tag/v0_38