fix(sdk/metric): Fix data race with initialization of NewPeriodicReader

[CTrando]

Fixes #8769

Just have to move the initialization of a field before the goroutine starts up. Test works by creating a bunch of workers with a small timeout on the goroutine to replicate the data race.

You have to run the test with the -race flag if you want to reproduce the data race test failure.

[linux-foundation-easycla]

The committers listed above are authorized under a signed CLA.

• ✅ login: CTrando / name: Cameron Trando (81fadb8)

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 89.1%. Comparing base (c80f83a) to head (9af0045).

Additional details and impacted files

@@           Coverage Diff           @@
##            main   #9028     +/-   ##
=======================================
- Coverage   89.1%   89.1%   -0.1%     
=======================================
  Files        338     338             
  Lines      22274   22274             
=======================================
- Hits       19859   19857      -2     
- Misses      2415    2417      +2     

Files with missing lines	Coverage Δ
sdk/metric/periodic_reader.go	87.4% <100.0%> (-0.6%)	⬇️

... and 3 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[ps-mir]

The fix looks ok. But test is not a reliable regression guard, specially when cpu contention is high.

Is there a way observ.NewInstrumentation can be held long enough. TestPeriodicReaderInstrumentationError could be of some use.

[ps-mir]

Much better, This test is deterministic in producing the Reader non registration and fails 60/60 without fix.

A efficiency nit for reducing time.After.