Don't implicitly trust the monotonic clock when calculating span end times

[kstenerud]

Goal

We're getting occasional broken end times that are theorised to be caused by skews in the monotonic clock. This PR adds code to validate any time values calculated using the monotonic clock to ensure they're not too far out from the regular clock-based end time.

Testing

Added more unit tests.

[github-actions]

    FILE SIZE        VM SIZE    
 --------------  -------------- 
  +0.1%    +112  +0.1%    +112    String Table
  +0.1%     +80  +0.1%     +80    Symbol Table
  +0.1%     +60  +0.1%     +60    __TEXT,__text
  +0.7%     +16  +0.7%     +16    __TEXT,__const
  +0.2%      +8  +0.2%      +8    __TEXT,__unwind_info
  +0.0%      +1  +0.0%      +1    __TEXT,__objc_methname
  -0.1%      -8  -0.1%      -8    __TEXT,__gcc_except_tab
  -0.2%     -77  -0.2%     -77    [__TEXT]
  [ = ]       0 -51.1%    -192    [__LINKEDIT]
  +0.0%    +192  [ = ]       0    TOTAL

Generated by 🚫 Danger