ci: shard + cache for faster runs

[Anmol1696]

Summary

Reduces CI wall-clock time and runner cost across three workflows by eliminating duplicate work, parallelising sequential steps, and skipping CI on non-code changes.

What changed and why

1. Remove redundant fn-init-e2e job (test.yaml) — saves 1 runner per trigger

The fn-init-e2e job ran pnpm exec jest tests/integration/fn-init.test.ts — exactly the same test file that the integration job already runs as part of pnpm test:integration (--testPathPatterns='tests/integration'). It also duplicated the six-package toolkit build that the toolkit job performs. Every CI trigger was spinning up a fourth parallel runner to do no new work.

• Measured timing (runs 26075627160, 26071351252): fn-init-e2e consistently completed in 44–52 s.
• Expected saving: ~44–52 s of parallel compute eliminated per trigger; wall-clock unchanged (other jobs are the bottleneck).

2. Split Build & Lint into parallel build + lint jobs (ci.yaml) — saves ~7–10 s wall-clock

Previously a single sequential job ran: generate → install → build (22 s) → lint (~8 s). Lint was blocked behind the full build even though it doesn't need build artifacts.

Splitting into two independent parallel jobs each doing their own generate + install means the wall-clock is now max(build, lint) ≈ 46 s instead of build + lint ≈ 53 s.

• Measured timing (step timestamps from run 26071351252): pnpm build step took 22.3 s; pnpm lint step is independent.
• Expected saving: ~7–10 s off the CI workflow's wall-clock per trigger.

3. paths-ignore on ci.yaml, test.yaml, docker.yaml — full skip on docs/k8s-only PRs

All three workflows previously ran on every PR regardless of what changed. Changes to docs/**, k8s/**, or markdown files don't affect build, lint, unit tests, integration tests, toolkit tests, or Docker images.

• ci.yaml + test.yaml: ignore docs/**, k8s/**, **/*.md.
• docker.yaml: additionally ignores tests/** (changing tests doesn't affect what images are built). Note: docker.yaml already rebuilds only images whose source changed via matrix; this filter handles the "nothing to rebuild" case.
• test-k8s-deployment.yaml already had fine-grained paths: filters — untouched.
• Expected saving: full skip of ~50 s build + ~52 s tests + ~2 min Docker on docs/k8s-only PRs.

4. --frozen-lockfile consistency (test.yaml)

ci.yaml already used pnpm install --frozen-lockfile; test.yaml did not. Added the flag to all three pnpm install calls in test.yaml so CI can't silently succeed with a stale lockfile.

What was NOT changed

• No product code, test code, or package versions touched.
• test-k8s-deployment.yaml left intact — already has path filters and per-function matrix parallelism.
• docker.yaml build matrix and GHA layer cache (cache-from: type=gha) left intact.
• actions/setup-node@v4 pin in docker.yaml preserved (intentional — comment explains v5 breaks when pnpm store is absent).

Test plan

• Verify CI and Tests workflows trigger on a code-only PR and all jobs pass.
• Verify CI and Tests workflows are skipped on a docs-only PR (e.g., a README.md change).
• Confirm tests/integration/fn-init.test.ts still runs (covered by integration job's pnpm test:integration).
• Check that build and lint jobs in CI run concurrently, not sequentially.

🤖 Generated with Claude Code

[Copilot]

Pull request overview

This PR optimizes GitHub Actions CI execution by skipping workflows on non-code-only changes, removing redundant test execution, and parallelizing build vs lint to reduce wall-clock time and runner usage.

Changes:

• Add paths-ignore filters to ci.yaml, test.yaml, and docker.yaml to skip CI runs for docs/k8s/markdown-only changes (and tests-only changes for Docker publishing).
• Split the CI “Build & Lint” job into separate parallel build and lint jobs.
• Remove the redundant fn-init-e2e job and enforce pnpm install --frozen-lockfile consistently in test.yaml.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File	Description
.github/workflows/test.yaml	Adds path-based skipping, enforces frozen lockfile installs, and removes duplicate fn init e2e job.
.github/workflows/docker.yaml	Adds path-based skipping (including tests/**) to avoid unnecessary image builds on non-image-affecting changes.
.github/workflows/ci.yaml	Adds path-based skipping and splits build/lint into parallel jobs for faster CI wall-clock.

---

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