Reduce integration test CI time: fewer restarts, one shared database

[Hazel-Datastax]

What this PR does:
This PR decreases the time of the integration test jobs. It makes fewer restarts, and most restarts keep the database container. It changes only test code and pom.xml.

Changes

1. Class order: TestResourceGroupingClassOrderer

• A new JUnit ClassOrderer sorts the integration test classes:
  1. The classes that only use DseTestResource.
  2. The other classes, grouped by their test resources.
• pom.xml sets the orderer with quarkus.test.class-orderer in the failsafe configuration. Surefire (unit tests) does not use it.
• QuarkusTestProfileAwareClassOrderer calls this orderer first and keeps its order for integration tests.
• The orderer finds the test resources on the class, its superclasses and its enclosing classes, as Quarkus does. It ignores global resources, because they apply to all classes.
• Result: 10 starts instead of 16. There is 1 start for the shared resources and 1 start for each of the 9 other classes.

2. Shared database container: StargateTestResource

• All test classes in the test JVM use one database container. A static field keeps it.
• stop() does not stop the shared container. Testcontainers removes it when the JVM stops.
• At each restart, the resource drops the keyspaces that earlier classes left. It keeps only the keyspaces that the new container had. The app starts on a clean database, as before.
• A resource can override needsDedicatedContainer() to get a container of its own. These test classes use this:
  • CreateCollectionTooManyIndexesIntegrationTest and CreateCollectionTooManyTablesIntegrationTest. Their limits count all collections and indexes in the database.
  • SessionEvictionIntegrationTest. The test stops and starts the database container.
• When the app runs in a container (containerNetworkId is set), the behavior does not change. Each start of the app gets a new network, so the resource starts a new database container each time. CI does not use this mode.
• Result: 4 database starts instead of 16. The other 6 restarts only restart the app.

Results

• Before: the first run on main after test: cover successful findAndRerank through the Nvidia HTTP client #2605 (run 37555686049). It has the same tests as this PR.
• After: this PR (run 37573527820).

	hcd-it before	hcd-it after	dse69-it before	dse69-it after
Job duration	33.1 min	24.4 min	30.2 min	24.8 min
App starts	16	10	16	10
New database containers	16	4	16	4
Stop and database start (total)	768 s	243 s	479 s	194 s
App start (total)	61 s	47 s	96 s	65 s
Tests (skipped)	1532 (4)	1532 (4)	1524 (66)	1524 (66)

• hcd-it is 8.7 min faster (−26%). dse69-it is 5.4 min faster (−18%).
• Compared with the mean of the last 20 successful runs on main, hcd-it decreased from 32.8 to 24.4 min (−26%), and dse69-it from 27.6 to 24.8 min (−10%). 19 of these 20 runs are before test: cover successful findAndRerank through the Nvidia HTTP client #2605. They have fewer tests and only 12 starts.
• A restart that keeps the database takes 10–11 s on HCD and 13–15 s on DSE, with the cleanup. Before, a restart took approximately 52 s on HCD and 36 s on DSE.
• Job durations change much from run to run. On main, hcd-it took 30.1–39.7 min in the last 20 runs. The number of starts and the restart times are more stable.

How to check

In the log of an integration test job:

• Started a new database container occurs 4 times.
• Reusing the shared database container occurs 6 times.
• The 78 classes that only use DseTestResource run first. The 9 other classes run after them.

Which issue(s) this PR fixes:
Fixes #2621

Checklist

• Changes manually tested
• Automated Tests added/updated
• Documentation added/updated
• CLA Signed: DataStax CLA

[github-actions]

➡️ Unit Test Coverage Delta vs Main Branch

Metric	Value
Main Branch	54.71%
This PR	54.71%
Delta	⚪ 0.00%
ℹ️ Coverage unchanged

[github-actions]

Unit Test Coverage Report

Overall Project	54.71%	🍏

There is no coverage information present for the Files changed

[github-actions]

➡️ Integration Test Coverage Delta vs Main Branch (hcd-it)

Metric	Value
Main Branch	74.27%
This PR	74.27%
Delta	⚪ 0.00%
ℹ️ Coverage unchanged

[github-actions]

Integration Test Coverage Report (hcd-it)

Overall Project	74.27%	🍏

There is no coverage information present for the Files changed

[github-actions]

📉 Integration Test Coverage Delta vs Main Branch (dse69-it)

Metric	Value
Main Branch	72.98%
This PR	72.95%
Delta	🔴 -0.04%
⚠️ Coverage decreased

[github-actions]

Integration Test Coverage Report (dse69-it)

Overall Project	72.95%	🍏

There is no coverage information present for the Files changed

[erichare]

LGTM @Hazel-Datastax ! I left one comment about how we handle the drop keyspaces, but otherwise this looks good and seems to be a huge improvement