benchmarks: add dependency-heavy package workload

[sonukapoor]

Summary

Add a narrow dependency-heavy/package-execution benchmark workload and wire it into the benchmark runner.

This PR adds:

• benchmarks/workloads/dependency-heavy-package.js
• a small local package-like fixture graph under benchmarks/workloads/fixtures/dependency-heavy-package/
• run.sh entry for dependency-heavy-package
• benchmarks/README.md notes for what this workload isolates and does not measure

What it measures

• module-loader plus execution cost on a multi-file dependency graph
• package-style code paths beyond startup-only micro workloads
• short-lived execution where import graph shape matters

What it does not measure

• third-party package install or registry/network fetch cost
• very large real-world dependency trees
• framework-level startup and steady-state behavior

Verification

./build-edge/edge benchmarks/workloads/dependency-heavy-package.js
node benchmarks/workloads/dependency-heavy-package.js
bun benchmarks/workloads/dependency-heavy-package.js
deno run benchmarks/workloads/dependency-heavy-package.js

WARMUP=5 RUNS=30 ./benchmarks/run.sh dependency-heavy-package

All runtimes produced the same checksum locally for the direct workload command.

Local checkpoint (WARMUP=5, RUNS=30)

• edge: 60.8 ms
• node: 38.6 ms
• bun: 19.5 ms
• deno: 30.8 ms

This is a narrow surface checkpoint and not a broad runtime verdict.