W-23599769: benchmark existing artifacts and CLI commands

[andres-rad]

Summary

• Allow Node, Python, and CLI benchmarks to use existing artifacts without local native builds.
• Attribute wrapper benchmark results to the native library selected by the external artifact.
• Replace the synthetic CLI harness with end-to-end dw run first-run measurements.

Validation

• ./gradlew native-cli:test
• ./gradlew native-lib:benchmarkJsUnitTest -PskipNodeTests=false
• ./gradlew native-lib:test -PskipNodeTests=true -PskipPythonTests=true using GraalVM 21.0.11
• Gradle dry runs for default and artifact-override benchmark task dependencies
• Ordinary CLI UTF-16 XML corpus command

Using Custom Artifacts

Runner	Variable	Expected artifact
Node wrapper	DW_BENCH_NODE_PACKAGE	Extracted @dataweave/native package containing dist/index.js and native/dwlib.*
Python wrapper	DW_BENCH_PY_SITE	Site-packages-style directory containing dataweave/__init__.py and dataweave/native/dwlib.*
CLI	DW_BENCH_BIN	Executable ordinary dw binary

DW_BENCH_NODE_PACKAGE=/tmp/artifacts/node/package \
DW_BENCH_PY_SITE=/tmp/artifacts/python/site-packages \
DW_BENCH_BIN=/tmp/artifacts/cli/dw \
./gradlew benchmarkCompare -Pbenchmark=true

When an override is set, Gradle skips that runner's corresponding local native artifact build. The engine runner remains local because it runs against this checkout's Maven runtime dependency.