metrics: fix TestExpDecaySampleNanosecondRegression sometimes failed

[Halimao]

TestExpDecaySampleNanosecondRegression will fail sometimes
​
Found this issue in ci job

--- FAIL: TestExpDecaySampleNanosecondRegression (0.00s)
    sample_test.go:144: out of range [14, 16]: 13.9

​
It hasn't always failed, I ran it on my local machine, and it's mostly been passed.
​
The problem is the test case using 100 as sample reservoirSize, which is a bit small.
​

• If we use 100 as reservoirSize and run the test 1000 times, it will most likely be failed.
• If we use 1000 as reservoirSize and run the test 10000 times, it will most likely be successful, I haven't failed at least

[rjl493456442]

LGTM.

I ran the stress test locally and it can indeed capture the failure after running 2minutes.

/var/folders/sm/ytlmy_n92dnc2sxp8c5g9bhr0000gn/T/go-stress-20240524T151135-281253435
--- FAIL: TestExpDecaySampleNanosecondRegression (0.00s)
    sample_test.go:144: out of range [14, 16]: 16.1
FAIL


ERROR: exit status 1

1m40s: 136 runs so far, 1 failures (0.74%)

/var/folders/sm/ytlmy_n92dnc2sxp8c5g9bhr0000gn/T/go-stress-20240524T151135-3139478043
--- FAIL: TestExpDecaySampleNanosecondRegression (0.00s)
    sample_test.go:144: out of range [14, 16]: 13.9
FAIL

After switching to the new parameters, it can run without issue.

4m45s: 400 runs so far, 0 failures
4m50s: 408 runs so far, 0 failures
4m55s: 408 runs so far, 0 failures
5m0s: 416 runs so far, 0 failures
5m5s: 424 runs so far, 0 failures
5m10s: 432 runs so far, 0 failures
5m15s: 440 runs so far, 0 failures
5m20s: 448 runs so far, 0 failures