Work on #1685

Improve unfiltered /timeseries read performance + perf coverage

• move the main unfiltered /timeseries read path off the Oracle retrieve_ts_out_tab hot path and onto direct AV_TSV_DQU/jOOQ reads while preserving retrieve_ts response semantics in Java
• keep the direct path aligned with Oracle behavior using integration parity coverage, including a DST-window regression case for regular series
• Java/Gradle benchmark harness
• add explicit fixture teardown for the benchmark runner so local perf runs exit cleanly and leave the workspace in a predictable state

Benchmark

Local single-request benchmarks against the Docker/Testcontainers Oracle dev environment, using page-size equal to the requested row count:

• legacy path, 250000 points: 40.679s
• new direct path, 250000 points: 4.652501s
• legacy path, 500000 points: HTTP 500

The current Java benchmark harness also confirms the direct path now returns the expected count for the large seeded series:

• direct path, 250000 points: 250000 reported rows, correct final timestamp, 4.652501s

The benchmark task writes JSON artifacts under load_data/performance/results/.

Validation

• ./gradlew :cwms-data-api:integrationTests --tests cwms.cda.api.TimeSeriesDirectReadParityIT
• ./gradlew :cwms-data-api:timeseriesReadBenchmark -Pbenchmark.pointCount=250000 -Pbenchmark.runs=1 -Pbenchmark.forceReseed=true

Notes

• scope here is the main unfiltered /timeseries read path; /timeseries/filtered is still on the legacy path
• could move to larger streaming/lower-materialization for more improvements

Seeding 250,000 TS values for SPK and reading them back via CDA returns in 4.6s

Before changes

Current 500k+ Fails
Current 250k in 40.679s

Changes

250k in 4.793s
500k in 7.498s
1,000,000 in about 14.9s