[fix][test] Made ProtobufNativeSchemaTest.testSchema order-independent

[LucasEby]

Fixes #24804

Motivation

In ProtobufNativeSchemaTest.testSchema, the json's contents and the FileDescriptorSet field's contents do not have a deterministic order but the hardcoded string assertion assumes a deterministic order. The json serializer did not guarantee attribute order and inside FileDescriptorSet the contents can also be in different orders due to different generation paths or environments producing the contents in different orders despite the logical content being the same. Since the original test compared the raw strings/trees "as-is", harmless re-ordering could flip the test from pass to fail without any real schema change.

Modifications

We no longer compare raw strings/trees "as-is". Instead the fileDescriptorSet field is base64-decoded into a FileDescriptorSet, then converted into a JSONObject. We compare the expected and actual descriptors with JsonAssert in NON_EXTENSIBLE mode, which ignores field ordering but forbids missing or extra fields.

Next, we remove FileDescriptorSet from the outer JSON and compare the remainder in the NON_EXTENSIBLE mode again. We implemented this two stage approach because JSON assertion does not re-order the inner contents of the FileDescriptorSet (compares directly as strings instead) even if though it's decoded contents can be re-ordered.

By decoding and normalizing fileDescriptorSet first, we ensure the test remains robust to nondeterministic ordering inside the descriptor set, while still validating the schema structure strictly. This change keeps the spirit of the original test while eliminating failures caused solely by allowed reordering.

NOTE: Some additional assertTrue and assertFalse assertions were also added to ensure the test remains robust in the event that the hardcoded string is ever modified for some reason. This will prevent someone from modifying the string and then forgetting to modify the test afterwards, thus causing the fileDescriptorSet assertion to incorrectly pass from comparing empty strings.

Verifying this change

• Make sure that the change passes the CI checks.

This change is already covered by existing tests, such as org.apache.pulsar.client.impl.schema.ProtobufSchemaTest#testSchema.

Does this pull request potentially affect one of the following parts:

If the box was checked, please highlight the changes

• Dependencies (add or upgrade a dependency)
• The public API
• The schema
• The default values of configurations
• The threading model
• The binary protocol
• The REST endpoints
• The admin CLI options
• The metrics
• Anything that affects deployment

Documentation

• doc
• doc-required
• doc-not-needed
• doc-complete

Matching PR in forked repository

PR in forked repository: LucasEby#1

[LucasEby]

@lhotari I incorporated your suggestions and made the code easier to follow. All checks passed on my fork: LucasEby#1
The code should be good to merge

[lhotari]

It could be easier to understand the logic if "fileDescriptorSet" would be ignored in the comparison of the schema json and compared separately. Assertions would optimally have a description message to explain what went wrong.

[LucasEby]

@lhotari I incorporated the requested changes. The changes passed CI checks on my personal fork: LucasEby#1

I additionally added some additional assertTrue and assertFalse assertions to ensure the test remains robust in the event that the hardcoded string is ever modified for some reason. The assertions ensure the retrieved fileDescriptorSet field contains data and the strings compared later won’t be empty, thus preventing the possible scenario of the fileDescriptorSet assertions incorrectly passing from comparing empty strings.