use string subject IDs in PRE_DEFINED_EXTRA_FIELDS_READ_GRANT_OBJECT …

[hu-ahmed]

Summary

• Fix: ThingEventEnricher now produces PRE_DEFINED_EXTRA_FIELDS_READ_GRANT_OBJECT
  header with string subject IDs instead of integer indices, matching the format
  expected by DittoCachingSignalEnrichmentFacade
• Add ReadGrant.toJson() to serialize grants as {"/path": ["subject1", "subject2"]}
• Remove dead code: IndexedReadGrant and ReadGrantIndexer (no remaining usages)
• Add regression tests verifying string format in both ReadGrantCollectorTest
  and ThingEventEnricherTest

Root cause

ReadGrantIndexer.index() converted ReadGrant to an indexed format where
subject IDs are replaced by integer indices to reduce header size:

{"/attributes": [0]}                                      

Where 0 maps to a subject ID string. But
DittoCachingSignalEnrichmentFacade.filterAskedForFieldSelectorToGrantedFields()
compares the array values directly against the requesting connection's authorization
subject ID strings. JsonValue(integer) never equals JsonValue(string),
so the filter returned an empty field selector, causing all pre-defined extra fields
to be silently dropped from enriched events.

Impact

All connections relying on pre-defined extra fields enrichment (e.g.
extraFields=attributes) received events with empty extra, causing
downstream consumers to fail when attempting to read expected fields.

[thjaeckle]

@hu-ahmed thanks for the fix - just a minor comment to adjust

[thjaeckle]

Thanks @hu-ahmed LGTM 👍