refactor(config): merge test config files

[317787106]

What does this PR do?

Consolidates test configuration files by deleting four redundant config files (args-test.conf, config-test-index.conf, config-test-mainnet.conf, config-test-storagetest.conf) and standardizing tests to use a single shared config-test.conf. Also renames config-localtest.conf to config-shield.conf to better reflect its purpose (shielded-transfer tests), and removes the hardcoded seed node list from config-test.conf since unit tests have no need for P2P connectivity.

Specifically:

• Delete 4 redundant test config files (~940 lines removed)
• Rename config-localtest.conf → config-shield.conf
• Remove NET_CONF, MAINNET_CONF, LOCAL_CONF, STORAGE_CONF, INDEX_CONF from TestConstants; add SHIELD_CONF
• Update all affected test classes to reference TestConstants.TEST_CONF or TestConstants.SHIELD_CONF instead of bare string literals
• Clear seed node IPs from config-test.conf; adjust ArgsTest seed-node assertion from 11 → 0

Why are these changes required?

The previous test suite maintained five separate .conf files that largely duplicated each other, making it hard to know which config to use for a new test and causing drift over time (e.g., config-test-mainnet.conf contained mainnet genesis/seed data that is irrelevant and potentially misleading in a unit-test context). Centralizing to one default test config (config-test.conf) and one purpose-specific config (config-shield.conf) reduces maintenance overhead and eliminates accidental coupling to live-network parameters in tests.

This PR has been tested by:

• Unit Tests

Extra details

StorageTest gains additional coverage to compensate for the removal of config-test-storagetest.conf-dependent tests that were folded into the unified config.

[barbatos2011]

The reflection-based setPrivateField here bypasses the explicit @Setter(AccessLevel.NONE) on StorageConfig.defaultDbOption / defaultMDbOption / defaultLDbOption. Those annotations document an encapsulation contract — those fields are bound by ConfigBeanFactory and should be immutable post-binding — which this code silently breaks.

Three concrete downsides:

1. Refactor-fragile: the field name is a string literal ("defaultDbOption"). A future field rename via IDE will not touch this string, the project still compiles, and the test only fails at runtime with NoSuchFieldException.
2. Integration coverage lost: the deleted config-test-storagetest.conf exercised the full ConfigFactory → ConfigBeanFactory → @Setter(NONE) fields path. Bypassing it with reflection turns this into a unit test of utility functions; a future regression in ConfigBeanFactory binding of @Setter(NONE) fields would not be caught here.
3. Anti-pattern propagation: once setPrivateField exists in the test tree, future tests are likely to copy it instead of pursuing cleaner solutions.

Suggested alternative — use ConfigFactory.parseString to exercise the real binding path with zero reflection and refactor-safety:

private static void setupStorage() {
  Config cfg = ConfigFactory.parseString(
      "storage.defaultDbOption.maxOpenFiles = 50\n" +
      "storage.defaultMDbOption.maxOpenFiles = 500\n" +
      "storage.defaultLDbOption.maxOpenFiles = 1000\n" +
      "storage.properties = [\n" +
      "  { name = account, path = storage_directory_test, maxOpenFiles = 100, ... },\n" +
      "  { name = account-index, path = storage_directory_test, maxOpenFiles = 100, ... },\n" +
      "  { name = test_name, path = test_path, createIfMissing = false, blockSize = 2, ... }\n" +
      "]")
      .withFallback(ConfigFactory.defaultReference());
  StorageConfig sc = StorageConfig.fromConfig(cfg);
  storage.setDefaultDbOptions(sc);
  storage.setPropertyMapFromBean(sc.getProperties());
}

If @Setter(NONE) truly needs to be relaxed for testability, prefer @Setter(AccessLevel.PACKAGE) or an explicit @VisibleForTesting setter on StorageConfig — both are type-safe and document the intent.

[317787106]

Good suggestion, use the real config string instead.

[barbatos2011]

Passing "" as the second argument here is a code smell — Configuration.getByFileName explicitly rejects blank input with IllegalArgumentException("Configuration path is required!"), so an empty string is documented as invalid.

The test currently works because the -c TestConstants.TEST_CONF flag short-circuits the fallback path (in Args.setParam line ~166: isNoneBlank(shellConfFileName) ? shellConfFileName : confFileName). But this leaves a fragile dependency: if a future refactor removes the -c flag from this call site (e.g. switching to a different mechanism for selecting the test config), the empty-string second arg silently flows into getByFileName and the test fails with an opaque IllegalArgumentException.

Two safer options:

1. Keep TestConstants.NET_CONF as the fallback constant (re-add it, semantically harmless since the body of the new config-test.conf covers what NET_CONF used to point at).
2. Add an Args.setParam(String[]) overload that resolves confFileName internally (or accepts Optional<String>) — eliminates the misleading placeholder pattern across the test suite.

Not blocking; just worth tracking before this pattern gets copy-pasted into other tests.

[317787106]

Thanks for your suggestion, use the standard method instead although it doesn't matter much:

Args.setParam(new String[] {"--keystore-factory"}, TestConstants.TEST_CONF);

[barbatos2011]

Follow-up observation, not blocking: although this PR deletes four redundant config files (-996 lines, well-deserved cleanup), the consolidated config-test.conf itself drops only 18 lines (387 → 369), and the renamed config-shield.conf drops 17. So the consolidation eliminates file drift risk but doesn't reduce per-test config-load surface — each of the ~21 affected tests still loads a 369-line conf, mostly boilerplate it doesn't read.

That's a real, but smaller, win. Worth being explicit in the PR description:

Note: this PR consolidates files but does not slim the unified config-test.conf (~369 lines) itself. Per-test inline overrides via ConfigFactory.parseString are a follow-up.

Otherwise reviewers (and future contributors) may expect the consolidation step alone to address the boilerplate problem, when in fact the boilerplate just got concentrated in one place.

Concrete follow-up path, if someone wants to pursue it later: tests with narrow needs (e.g. AssetUpdateHelperTest, StorageTest, ApiUtilTest) can move to inline ConfigFactory.parseString(...).withFallback(ConfigFactory.defaultReference()) and shed their dependency on TEST_CONF entirely. reference.conf already provides the safety net thanks to PR #6615.

[halibobo1205]

[NIT] docs still mention deleted test resource config-localtest.conf in implement-a-customized-actuator-{en/cn}.md