refactor(tracking): replace flat operationType with dimensional attributes

[jphui]

Summary

• Rewrite BaseDaoBenchmarkMetrics so impls receive each dimension (operation, aspect, count bucket, status, error class) as a separate parameter instead of one concatenated operationType string.
• Update InstrumentedEbeanLocalAccess to pass dimensions through cleanly: aspect class name for per-aspect ops, a bucketed count label (1..9, 10+) for batch ops.
• Bucketing keeps cardinality bounded for downstream backends (OTEL especially).

Why

Today the impl gets strings like add.AliasesInfo, create.aspects_3, batchGetUnion.keys_42 — each becomes its own metric path in RRD/Dropwizard, so aggregating "all adds across all aspects" requires summing hundreds of named series. Splitting into native dimensions lets backends emit one metric with attributes and aggregate/filter natively. This is a pre-req for a downstream OTEL-backed impl.

Follow-up (required in metadata-graph-assets)

This PR changes the interface contract; it does not itself emit any metrics. The existing DropwizardDaoBenchmarkMetrics in metadata-graph-assets (the only known concrete impl) implements the old recordOperationLatency / recordOperationError methods and will not compile against the new contract.

When MGA picks up this version (via a metadata-models bump), it must in the same PR:

1. Replace DropwizardDaoBenchmarkMetrics with a new DaoBenchmarkOtelMetrics that implements the single recordOperation(...) method.
   • Mirror the existing MgaClientMetrics facade pattern (typed AttributeKeys, pre-built LongCounter + LongHistogram, Meter obtained from com.linkedin.opentelemetry.OpenTelemetry.meterBuilder(...)).
   • Suggested metric names: Dao.Benchmark.Operation.Count (counter), Dao.Benchmark.Operation.Latency (histogram, ms).
   • Attribute keys (PascalCase, per LinkedIn MDM convention): Operation, EntityType, Aspect, CountBucket, Status, ErrorClass.
2. Update DaoBenchmarkMetricsFactory.createInstance(...) to return the OTEL impl instead of the Dropwizard one.
3. Delete DropwizardDaoBenchmarkMetrics and its DropwizardDaoBenchmarkMetricsTest.
4. Add unit tests using InMemoryMetricReader (mirror MgaClientMetricsTest).

The MGA-side OTEL impl is where the real metric emission lives — this PR is purely kernel plumbing.

Test plan

• :dao-api + :dao-impl:ebean-dao compile clean (Java 11)
• InstrumentedEbeanLocalAccessTest passes — rewrote assertions to verify the 7-param contract, added bucketCount boundary tests (0, 1, 5, 9, 10, 100), covered success + failure paths
• Full :dao-impl:ebean-dao:test suite (CI will run)

🤖 Generated with Claude Code

[codecov-commenter]

Codecov Report

❌ Patch coverage is 80.00000% with 6 lines in your changes missing coverage. Please review.
✅ Project coverage is 66.86%. Comparing base (b025046) to head (26f992b).

Files with missing lines	Patch %	Lines
...din/metadata/dao/InstrumentedEbeanLocalAccess.java	80.00%	6 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff              @@
##             master     #623      +/-   ##
============================================
+ Coverage     66.82%   66.86%   +0.03%     
- Complexity     1880     1885       +5     
============================================
  Files           148      148              
  Lines          7262     7273      +11     
  Branches        879      880       +1     
============================================
+ Hits           4853     4863      +10     
- Misses         2024     2025       +1     
  Partials        385      385              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.