Fix data race in periodic reader tests

[dashpole]

Fixes #4543

~/go/src/go.opentelemetry.io/opentelemetry-go/sdk/metric$ go test -race -count=1000 -timeout=2m -run "^TestPeriodicReaderRun$|^TestPeriodicReader$"
PASS
ok  	go.opentelemetry.io/otel/sdk/metric	46.140s

The Reader is already being initialized in the individual tests, e.g.

opentelemetry-go/sdk/metric/reader_test.go

Lines 56 to 57 in 47ba653

	func (ts *readerTestSuite) TestSDKProducer() {
	ts.Reader = ts.Factory()

The creation of an additional reader during registration causes problems because the extra reader was being overwritten in the test, and thus never properly Shutdown(). The run() goroutine from the reader was still active during other tests, and could race with them.

[codecov]

Codecov Report

Merging #4731 (7c38a0e) into main (47ba653) will decrease coverage by 0.1%.
The diff coverage is n/a.

Additional details and impacted files

@@           Coverage Diff           @@
##            main   #4731     +/-   ##
=======================================
- Coverage   81.8%   81.7%   -0.1%     
=======================================
  Files        224     224             
  Lines      18113   18113             
=======================================
- Hits       14817   14815      -2     
- Misses      3000    3002      +2     
  Partials     296     296             

see 1 file with indirect coverage changes

[pellared]

I think that some tests also miss shutdown of the reader which also can bring problems.
Can you also please double-check these tests?

• TestPeriodicReaderMultipleForceFlush
• BenchmarkPeriodicReader (it is a benchmark so it is usually not run, but still it can interfere if one runs the benchmarks)

[dashpole]

Added shutdowns to the two you mention. Since tests are run serially, it doesn't seem to matter, but good to do anyways.

before:

BenchmarkPeriodicReader/Collect-24 465 2940 ns/op 192 B/op 6 allocs/op

after:

BenchmarkPeriodicReader/Collect-24 546 2229 ns/op 192 B/op 6 allocs/op