Allow negative performance timings

[mhennoch]

Which problem is this PR solving?

• As mentioned in 1769 PerformanceNavigationTiming.fetchStart can be negative by a tiny amount ( seen values -1,-2, -3 during automated testing). Currently these entries get filtered out and no spans are created. We would like to see these spans and decide what to do with them in the backend.

Short description of the changes

• Remove value >= 0 checks from both entry sources (PerformanceNavigationTiming and legacy perf.timing ). If they are negative then nothing should break:
  • If PerformanceNavigationTiming.fetchStart is -1 then when being converted to hrTime it will just be perf.timeOrigin - 1 which is close to the correct timestamp.
  • Same with legacy timings, although I've never seen them being negative usually just out of order (We've seen responseEnd before fetchStart for example : ) ) or 0.
• Add and refactor tests. Replace Object.defineProperty usage with stubs(patch perf.timing getter) so that test order is not important anymore.

[codecov]

Codecov Report

Merging #330 (d987ff0) into main (dbbf02f) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##             main     #330   +/-   ##
=======================================
  Coverage   93.88%   93.88%           
=======================================
  Files          12       12           
  Lines         409      409           
  Branches       45       45           
=======================================
  Hits          384      384           
  Misses         25       25           

[johnbley]

👍