Fix benchmark workflow: drop goloader, fix proto panic, fix wazero module exit

[Copilot]

The weekly benchmark workflow was failing at update_readme.py with "No benchmark results found" because the test binary either failed to link (goloader) or panicked at startup (proto conflict) before producing any output.

goloader removed (Go 1.26.1 incompatible)

Go 1.26.1 rejects //go:linkname references to unexported runtime symbols at link time when the symbol is actually called. goloader calls runtime.itabTable, runtime.itabLock, runtime.firstmoduledata, runtime.gostringnocopy, and several others — all now blocked. No local patch can fix this without rewriting the library's core approach.

• Removed BenchmarkGoloaderRandInt and the github.com/pkujhd/goloader import from plugin_test.go
• Dropped the dependency from go.mod and removed the cp cmd/internal→cmd/objfile + go tool compile goloader.o Dockerfile steps
• Marked the README row N/A with an explanatory footnote; removed goloader from update_readme.py's pattern map

Protobuf registration panic fixed

github.com/GoCodeAlone/go-plugin and github.com/hashicorp/go-plugin both register identical proto file paths (internal/plugin/grpc_broker.proto, etc.), causing a panic in init() before any benchmark ran.

ENV GOLANG_PROTOBUF_REGISTRATION_CONFLICT=warn

Wazero benchmark fixed

runtime.Instantiate auto-runs _start, which in TinyGo's WASI target calls main() and then proc_exit(0), closing the module. Every RandInt call in the loop then fails with "module closed with exit_code(0)".

// Before: runtime.Instantiate runs _start → proc_exit(0) → module closed
module, err := runtime.Instantiate(ctx, wasmFile)

// After: skip _start so the module stays alive across benchmark iterations
cm, _ := runtime.CompileModule(ctx, wasmFile)
cfg := wazero.NewModuleConfig().WithStartFunctions() // no start functions
module, _ := runtime.InstantiateModule(ctx, cm, cfg)

---

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.

[Copilot]

Pull request overview

This PR fixes the weekly benchmark automation by removing the goloader benchmark/dependency (incompatible with Go 1.26.1), preventing a protobuf registration panic at startup, and adjusting the wazero benchmark so the WASM module doesn’t auto-exit before iterations run.

Changes:

• Remove goloader benchmark code and module dependency; update README and README-update script accordingly.
• Set GOLANG_PROTOBUF_REGISTRATION_CONFLICT=warn in the Docker benchmark environment to avoid init-time panics.
• Change wazero benchmark instantiation to skip _start so the module remains usable across benchmark iterations.

Reviewed changes

Copilot reviewed 6 out of 7 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
scripts/update_readme.py	Removes goloader from the benchmark-to-README row mapping to keep automation in sync.
plugin_test.go	Drops goloader benchmark/imports; updates wazero benchmark to compile + instantiate without auto-start.
go.mod	Removes the goloader module requirement.
go.sum	Removes goloader checksums.
README.md	Marks goloader as N/A with an explanatory footnote; updates benchmark numbers/rows.
Dockerfile	Adds protobuf conflict env var; removes goloader-specific build steps.
.github/workflows/benchmark.yml	Removes workflow comments related to the prior goloader workaround.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

In this benchmark, compile/instantiate failures are handled by printing and returning, which makes go test succeed while silently skipping the wazero sub-benchmark. That can reintroduce the "missing benchmark results" failure mode (or leave README rows stale). Prefer failing the benchmark (e.g., b.Fatalf / b.Fatal) so CI clearly reports the problem.

[Copilot]

Within the b.Run("wazero", ...) body, an error from randIntFunction.Call is currently handled by printing and returning, which ends the sub-benchmark early and can yield misleading results. Prefer failing the benchmark via the testing.B API (e.g., b.Fatalf / b.FailNow) so the run is marked unsuccessful instead of reporting partial numbers.