fix(openfeature): use delta temporality for OTLP metrics export

[sameerank]

What Does This Do

Changes the OTLP metrics exporter from alwaysCumulative() to deltaPreferred() for feature flag evaluation metrics.

Motivation

OTLP receivers (like the Datadog Agent) expect delta temporality for counter metrics. With cumulative temporality, each export sends the total count since start (5, 10, 15...), and receivers that sum these values produce overcounted results (5, 15, 30...).

With delta temporality, each export sends only the increment since the last export (5, 5, 5...), which receivers correctly sum to the actual total.

Testing in https://github.com/DataDog/ffe-dogfooding I saw Java was overcounting with OTLP, compared to the other SDKs.

This aligns with Datadog's OTLP documentation which recommends delta temporality for metrics.

Datadog works best with delta aggregation temporality for monotonic sums, histograms, and exponential histograms.

Additional Notes

• deltaPreferred() uses DELTA for Counter/Histogram and CUMULATIVE for UpDownCounter (per OTel spec)
• Updated test to use delta temporality to match production configuration
• FlagEvalMetrics is the only direct OTLP exporter in dd-trace-java; other OTel metrics go through the DogStatsD shim

Contributor Checklist

• Format the title according to the contribution guidelines
• Assign the type: and (comp: or inst:) labels in addition to any other useful labels
• Avoid using close, fix, or any linking keywords when referencing an issue
• Update the CODEOWNERS file on source file addition, migration, or deletion
• Update public documentation with any new configuration flags or behaviors

[dd-octo-sts]

Hi! 👋 Thanks for your pull request! 🎉

To help us review it, please make sure to:

• Add at least one type, and one component or instrumentation label to the pull request

If you need help, please check our contributing guidelines.