Name resolution delay

[AgraVator]

This finishes the remaining work in #12807

[kannanjgithub]

Can you check the missing code coverages from the report

./gradlew :grpc-api:test :grpc-api:jacocoTestReport :grpc-core:test :grpc-core:jacocoTestReport :grpc-opentelemetry:test :grpc-opentelemetry:jacocoTestReport

and fix them?

[kannanjgithub]

There are still some missing branches in codecov report. I have commented about missing coverages for OpenTelemetryMetricsModule.
Also address still existing coverages in the report except for cases where it cannot be done without changing class visibility or other reasons.

[kannanjgithub]

Branch getRealStream() == null && delayEnded == true is missing coverage. To cover it, a unit test only needs to invoke endDelay() directly on a pending stream while getRealStream() == null and then call updateDelay(...):

// Example in DelayedClientTransportTest.java
PendingStream pendingStream = ...;
pendingStream.endDelay(); // sets delayEnded = true while getRealStream() is still null
pendingStream.updateDelay("connecting", "new reason"); // triggers missing branch > returns

[kannanjgithub]

Lets not do this, I didn't realize this change requires changing the class access modifier for PendingStream.

[kannanjgithub]

The is reported missing coverage for the branch when activeDelayType is null but it cannot be meaningfully tested because determineQueuingDelayType always returns non null value and there are other complications if we make this unconditional.

[kannanjgithub]

When activeCallDelayStopwatch != null is true, invoking recordCallDelayStart with a different delayType while an existing call delay is already active to transition between delay segments is not exercised.

We can cover this branch by adding a transition call with a different delay type (e.g. "connecting") to clientCallDelayDuration_duplicateStart_ignoresSecondStart in OpenTelemetryMetricsModuleTest.java:

  @Test
  public void clientCallDelayDuration_duplicateStart_ignoresSecondStart() {
    OpenTelemetryMetricsResource resource = GrpcOpenTelemetry.createMetricInstruments(
        openTelemetryTesting.getOpenTelemetry().getMeterProvider().get("grpc-java"),
        ImmutableMap.of("grpc.client.call.delay.duration", true),
        false);
    OpenTelemetryMetricsModule module = new OpenTelemetryMetricsModule(
        new FakeClock().getStopwatchSupplier(), resource, emptyList(), emptyList());
    OpenTelemetryMetricsModule.CallAttemptsTracerFactory factory =
        new OpenTelemetryMetricsModule.CallAttemptsTracerFactory(
            module, "target:///", CallOptions.DEFAULT, method.getFullMethodName(),
            emptyList(), io.opentelemetry.context.Context.root());

    factory.recordCallDelayStart("resolving", "first start");
    // Duplicate call with same delay type is ignored
    factory.recordCallDelayStart("resolving", "second start");

    // Transition to a different delay type while active (exercises Objects.equals == false)
    factory.recordCallDelayStart("connecting", "transition to connecting");
    factory.recordCallDelayEnd();

    assertThat(openTelemetryTesting.getMetrics())
        .anySatisfy(
            metric - assertThat(metric)
                .hasName("grpc.client.call.delay.duration"));
  }

[kannanjgithub]

In the test for the null delay type branch, assert that no metric was recorded for null (e.g. assertThat(openTelemetryTesting.getMetrics()).extracting("name").doesNotContain("grpc.client.call.delay.duration") or verify that only points with valid non-null grpc.delay_type attributes exist).