test(profiling): stabilize Poisson sampling filter spec (#8659)

The `should cap endTime to now() if event endTime is in the future` test set
`nowValue = 1000`, which is only ~10x the samplingInterval of 100. The
filter's initial `nextSamplingInstant` is an exponential RV with mean 100, so
P(initial > 1000) = e^-10 ≈ 4.5e-5: the while loop in filter() would skip,
`currentSamplingInstant` would stay 0, and the first assertion would fail.

Raise `nowValue` to 100000 (P(initial > 100000) = e^-1000 ≈ 0). The
resetInterval still bounds the while loop to ~2 iterations, so the other
assertions remain tight.

Co-authored-by: Ruben Bridgewater <ruben@bridgewater.de>

Author: szegedi

packages/dd-trace/test/profiling/profilers/poisson.spec.js

@@ -131,7 +131,11 @@ describe('PoissonProcessSamplingFilter', () => {
       now,
     })
     const prevNextSamplingInstant = filter.nextSamplingInstant
-    nowValue = 1000
+    // nowValue must comfortably exceed the initial nextSamplingInstant, which is an
+    // exponential RV with mean = samplingInterval = 100. P(initial > 100000) = e^-1000,
+    // so the first assertion below is effectively never flaky. The resetInterval still
+    // bounds the while loop in filter() to ~2 iterations, keeping the other assertions tight.
+    nowValue = 100000
     const event = { startTime: 0, duration: 1e6 }
     filter.filter(event)
     assert.ok(