Compute bean raw types in a Set/HashSet before materializing them

[gsmet]

It allows to avoid performing very badly when we have lots of beans of the same type, for instance when you have a lot of Panache repositories for various entities.

We materialize things with Lists in the end, and they might be a bit more optimized than before for small non-singleton lists.

The downside is that we have a bit more allocations but from my testing, it wasn't affecting startup times.

Fixes #51468

We went from (see elements on the left side):

to

The rest of the ArC container initialization is mostly class loading.

[quarkus-bot]

Status for workflow Quarkus CI

This is the status report for running Quarkus CI on commit e29b86a.

Failing Jobs

Status	Name	Step	Failures	Logs	Raw logs	Build scan
❌	Native Tests - Windows support	Setup GraalVM	⚠️ Check →	Logs	Raw logs	🚧

You can consult the Develocity build scans.

---

Flaky tests - Develocity

⚙️ JVM Tests - JDK 21

📦 extensions/quartz/deployment

❌ io.quarkus.quartz.test.timezone.TriggerPrevFireTimeZoneTest.testScheduledJobs - History

• expected: <2025-12-09T16:00:35Z> but was: <2025-12-09T16:00:36Z> - org.opentest4j.AssertionFailedError

org.opentest4j.AssertionFailedError: expected: <2025-12-09T16:00:35Z> but was: <2025-12-09T16:00:36Z>
	at org.junit.jupiter.api.AssertionFailureBuilder.build(AssertionFailureBuilder.java:151)
	at org.junit.jupiter.api.AssertionFailureBuilder.buildAndThrow(AssertionFailureBuilder.java:132)
	at org.junit.jupiter.api.AssertEquals.failNotEqual(AssertEquals.java:197)
	at org.junit.jupiter.api.AssertEquals.assertEquals(AssertEquals.java:182)
	at org.junit.jupiter.api.AssertEquals.assertEquals(AssertEquals.java:177)
	at org.junit.jupiter.api.Assertions.assertEquals(Assertions.java:1145)
	at io.quarkus.quartz.test.timezone.TriggerPrevFireTimeZoneTest.testScheduledJobs(TriggerPrevFireTimeZoneTest.java:71)

---

⚙️ JVM Tests - JDK 21 Semeru

📦 extensions/micrometer-opentelemetry/deployment

❌ io.quarkus.micrometer.opentelemetry.deployment.compatibility.MicrometerTimedInterceptorTest.testTimeMethod_Uni - History

• Stream has no elements - java.lang.IllegalArgumentException

java.lang.IllegalArgumentException: Stream has no elements
	at io.quarkus.micrometer.opentelemetry.deployment.common.MetricDataFilter.lambda$lastReading$2(MetricDataFilter.java:213)
	at java.base/java.util.Optional.orElseThrow(Optional.java:403)
	at io.quarkus.micrometer.opentelemetry.deployment.common.MetricDataFilter.lastReading(MetricDataFilter.java:213)
	at io.quarkus.micrometer.opentelemetry.deployment.common.MetricDataFilter.lastReadingDataPoint(MetricDataFilter.java:231)
	at io.quarkus.micrometer.opentelemetry.deployment.compatibility.MicrometerTimedInterceptorTest.testTimeMethod_Uni(MicrometerTimedInterceptorTest.java:174)
	at java.base/java.lang.reflect.Method.invoke(Method.java:586)
	at io.quarkus.test.QuarkusUnitTest.runExtensionMethod(QuarkusUnitTest.java:532)

---

⚙️ MicroProfile TCKs Tests

📦 tcks/microprofile-lra

❌ org.eclipse.microprofile.lra.tck.TckRecoveryTests.testCancelWhenParticipantIsUnavailable - History

• Expecting the metric Compensated callback was called Expected: a value equal to or greater than <1> but: <0> was less than <1> - java.lang.AssertionError

java.lang.AssertionError: 
Expecting the metric Compensated callback was called
Expected: a value equal to or greater than <1>
     but: <0> was less than <1>
	at org.hamcrest.MatcherAssert.assertThat(MatcherAssert.java:20)
	at org.eclipse.microprofile.lra.tck.TckRecoveryTests.assertMetricCallbackCalled(TckRecoveryTests.java:210)
	at org.eclipse.microprofile.lra.tck.TckRecoveryTests.testCancelWhenParticipantIsUnavailable(TckRecoveryTests.java:195)
	at java.base/jdk.internal.reflect.NativeMethodAccessorImpl.invoke0(Native Method)

[mkouba]

It looks good. In terms of allocations, this could sometimes allocate more (new HashSet() + List.copyOf()) but I think that in most cases it will be fine.

[gsmet]

Yeah I did some allocation profiling after this change and it did allocate a bit more, which is not surprising. It wasn't that bad though.