feat: add GitHub Actions CI and reusable benchmark workflow

[giortzisg]

Summary

Phase 4 of the implementation plan — GitHub Action integration.

• .github/workflows/ci.yml — lint, test, and Docker build checks for all 11 containers (Python + Go + tools)
• .github/workflows/benchmark.yml — reusable workflow for SDK repos to call from their CI
• .github/workflows/test-benchmark.yml — smoke test for the benchmark workflow
• .github/actions/benchmark/action.yml — composite action wrapper
• lib/github.py — sticky PR comment posting via gh CLI with per-endpoint overhead tables
• bench.py — wired run command to lib.runner, added post-comment CLI command
• tests/test_github.py — 5 tests for comment formatting

Verified locally with act

• lint job passes
• test job passes (9/9 tests)
• docker-build job passes for all 11 matrix entries:
  • django-baseline, django-instrumented
  • fakerelay, loadgen, postgres
  • go-net-http-baseline, go-net-http-instrumented
  • go-gin-baseline, go-gin-instrumented
  • go-echo-baseline, go-echo-instrumented

Usage by SDK repos

jobs:
  bench:
    uses: getsentry/sdk-benchmarks/.github/workflows/benchmark.yml@main
    with:
      app: python/django
      sdk-version: "git+https://github.com/${{ github.repository }}@${{ github.head_ref }}"

Test plan

• All CI jobs verified with act locally
• pytest tests/ — 9/9 pass
• ruff check . clean

🤖 Generated with Claude Code

[cursor]

Cursor Bugbot has reviewed your changes and found 2 potential issues.

^{Bugbot Autofix is OFF. To automatically fix reported issues with Cloud Agents, enable Autofix in the Cursor dashboard.}

[sentry]

Bug: The run_benchmark function only outputs raw data, but format_comment expects computed summary statistics. This results in empty, useless benchmark reports being posted in PR comments.
_{Severity: CRITICAL}

Suggested Fix

Implement the statistical analysis layer that was left as a stub. This involves adding logic to lib/metrics.py, lib/report.py, and lib/compare.py to process the raw iteration data from run_benchmark() and generate the summary, overhead, and regression fields that format_comment() requires to produce a meaningful report.

Prompt for AI Agent

Review the code at the location below. A potential bug has been identified by an AI
agent.
Verify if this is a real issue. If it is, propose a fix; if not, explain why it's not
valid.

Location: bench.py#L26-L30

Potential issue: The `run_benchmark()` function writes raw iteration data to a JSON file
but does not compute the summary statistics (like `overhead`, `regression`, `endpoints`)
that the `format_comment()` function expects. Consequently, when `format_comment()`
reads this data, it finds no summary information. While the code avoids crashing by
using `.get()` with default values, the resulting PR comment is functionally useless,
containing empty tables and missing data. The core logic for statistical analysis,
intended for files like `lib/metrics.py` and `lib/report.py`, is completely missing,
breaking the data contract between the data generation and reporting steps.