Run short CI jobs on the standard runner

[rdimitrov]

Summary

Every job in PR Checks and Main build runs on the paid ubuntu-8cores-32gb runner, including jobs that finish in one to three minutes and spend most of that on checkout, toolchain setup and image pulls. Those steps do not scale with cores. The standard ubuntu-latest runner is free for this public repository and has 4 vCPUs and 16 GB.

Move the setup-dominated jobs to the standard runner and keep the large runner where cores or the critical path matter.

Moved to ubuntu-latest	Large-runner minutes today
Linting / Lint Go Code	2.2
Operator CI / Operator Tests	2.2
Operator CI / Build Operator	1.3
E2E Tests / MCP Conformance (proxy, vmcp)	1.6
E2E Tests Core: core, api-workloads, api-clients, api-registry, api-misc, llm	14.1
Total	~21 of ~100 per run

Kept on the large runner: Go tests, Helm chart tests, operator integration and E2E, the E2E binary build (root of the E2E critical path), and the proxy, mcp-protocol, middleware, lifecycle, mcp-run, network-isolation and vmcp shards.

Expected saving: about 21 large-runner minutes per run, roughly $650 to $700 per month across PR and main runs.

Type of change

• Other: CI/CD

Test plan

• actionlint on the three changed workflows with the repo's custom runner labels declared: clean.
• git diff --check: clean.
• Compare job durations on this PR's own run against a recent main run. The moved jobs are expected to take up to about twice as long individually, without extending the overall PR Checks wall time, since none of them is on the critical path.

Does this introduce a user-facing change?

No.

Special notes for reviewers

• Shard runner selection is runs-on: ${{ matrix.runner || 'ubuntu-8cores-32gb' }}, so shards without a runner field are unchanged. Moving a shard back is a one-line revert on that entry.
• The core shard runs Ginkgo with four processes; the standard runner's four vCPUs cover that.
• If a moved E2E shard turns out to be materially slower or flakier on the smaller runner, the intended fix is to move that one shard back, not to revert the PR.

Generated with Claude Code

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 78.70%. Comparing base (1e80f9d) to head (6fd88eb).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #6541      +/-   ##
==========================================
+ Coverage   78.69%   78.70%   +0.01%     
==========================================
  Files         777      777              
  Lines       76797    76797              
==========================================
+ Hits        60434    60443       +9     
+ Misses      16358    16349       -9     
  Partials        5        5              

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.