otlptracegrpc: Fix start/stop sync in mockCollector

[pellared]

Noticed when looking at #2686

This PR is intended to get rid of <-time.After(number * time.Millisecond) and properly synchronize the mock collector startup and shutdown. It should make the tests more stable.

For me before this fix running go test -race -count=100 is almost hanging (240sec for me). After this fix I get:

$ go test -race -count=100
--- FAIL: TestExporterShutdown (0.02s)
    --- FAIL: TestExporterShutdown/testClientStopHonorsTimeout (0.01s)
        client.go:73: expected context DeadlineExceeded error, got <nil>
FAIL
exit status 1
FAIL    go.opentelemetry.io/otel/exporters/otlp/otlptrace/otlptracegrpc 27.677s

$ go test -race -count=100
--- FAIL: TestExporterShutdown (0.03s)
    --- FAIL: TestExporterShutdown/testClientStopHonorsCancel (0.01s)
        client.go:92: expected context canceled error, got <nil>
FAIL
exit status 1
FAIL    go.opentelemetry.io/otel/exporters/otlp/otlptrace/otlptracegrpc 27.049s

(before this fix the failure is after 241.189s).

So not only it makes the tests more trustworthy (stable) but also shortens the execution time by 10 times (on my PC).