Bug 1880213 - Remove telemetry assertion that doesn't always hold r=sfink

This is another case where we occastionally get results that seem impossible
when dealing with time durations.

In this case we're already clamping the value before we send it to the
telemetry system, probably because we already ran into a case line this. The
telemetry we get out looks reasonable. Therefore I think it's fine to remove
this assertion.

Differential Revision: https://phabricator.services.mozilla.com/D201810

Author: jonco3

js/src/gc/Statistics.cpp

@@ -1134,7 +1134,6 @@ void Statistics::sendGCTelemetry() {
       uint32_t threadCount = gc->markers.length();
       double speedup = parallelMarkTime / wallTime;
       double utilization = parallelRunTime / (wallTime * threadCount);
-      MOZ_ASSERT(utilization <= 1.0);
       runtime->metrics().GC_PARALLEL_MARK_SPEEDUP(uint32_t(speedup * 100.0));
       runtime->metrics().GC_PARALLEL_MARK_UTILIZATION(
           std::clamp(utilization * 100.0, 0.0, 100.0));