Improve macOS timer resolution about php-spx

[orls]

I think that'll be usable for wall time, but the on-cpu time will still be microseconds multiplied up – not sure if you'd consider that a problem?

I'll have a go at this soon. I've been meaning to check out the new UI anyway. (It looks amazing, btw!)

[NoiseByNorthwest]

Damn you're right. I've just forgot the CPU time component. We need the same resolution for both timers otherwise this could lead to some disturbing inconsistencies.
However wall / cpu time inconsistency is theorycally handled. I will check the current behavior on Linux with 1ns wall time and 1us cpu time.

I've also seen here that some later macOS versions have clock_gettime support. According to statcounter.com macOS 10.12+ versions seem to have 60%+ market shares.

[orls]

Indeed (If you recall, in my initial impl I polyfilled it for < 10.12 because it was called globally, but then you consolidated all timing into spx_resource_stats-...).

It looks like it supports CLOCK_PROCESS_CPUTIME_ID too, which is great.

Personally I'd like to keep compatibility for 10.11, at least for now (colleagues & special snowflakes)...so how about I implement nsec resolution using clock_gettime for 10.12+ and fall back to usec for <= 10.11? The fallback mess will be isolated to only spx_resource_stats-macos.c.

[NoiseByNorthwest]

Yes, this is what I was thinking about.
I do not want to break compatibility with 10.11-, but we can do OS version dispatching (confined to spx_resource_stats-macos.c as you said or even subfiles if implementations are totally differents) to implement the optimal clock for each version.

[NoiseByNorthwest]

Fixed by #36