chore(ci): use non-large runners for release builds

[jdx]

Summary

• Drop -large from the Linux and macOS Namespace runner profiles used by the release tarball jobs in .github/workflows/release.yml.
• These jobs now use namespace-profile-endev-linux-amd64 and namespace-profile-endev-macos-arm64, matching what the rest of the workflows (test, registry, docs, etc.) already use.

Why

• Release builds don't need the larger instance profile, so this reduces CI cost for tag pushes without changing what gets built.

Notes

• hyperfine.yml still uses -large since it's a benchmarking workflow, not a release build — left untouched.

🤖 Generated with Claude Code

---

Note

Low Risk
Low risk CI-only change that just switches the runner profiles for the Linux and macOS release tarball jobs; main impact is potential runtime/performance differences on smaller runners.

Overview
Updates the release GitHub Actions workflow to run the Linux and macOS tarball build jobs on the standard Namespace runner profiles by removing the -large suffix.

This reduces resource usage/cost for release builds without changing build steps or artifacts.

^{Reviewed by Cursor Bugbot for commit 2535660. Bugbot is set up for automated code reviews on this repo. Configure here.}

[gemini-code-assist]

Note

Gemini is unable to generate a review for this pull request due to the file types involved not being currently supported.

[greptile-apps]

Greptile Summary

Two Namespace CI runner profiles for release tarball builds are switched from their -large variants to the standard profiles already used by the rest of the repo's workflows, reducing CI cost for tag pushes.

• build-tarball-linux now runs on namespace-profile-endev-linux-amd64 (was …-large); existing 45-minute job/retry timeout is unchanged.
• build-tarball-macos now runs on namespace-profile-endev-macos-arm64 (was …-large); existing 90-minute job/retry timeout is unchanged.
• hyperfine.yml benchmarking workflow retains its -large runner, as noted in the PR description.

Confidence Score: 5/5

Straightforward CI cost-reduction change with no risk to release artifact correctness or build logic.

The only change is dropping -large from two runner profile strings. The same profiles are already proven to work for other workflows in this repo, so the risk of these builds failing due to resource constraints is low. The existing generous timeouts (45 min Linux, 90 min macOS) with 3-attempt retries provide a cushion if the non-large runners are modestly slower on a given run.

No files require special attention.

Important Files Changed

Filename	Overview
.github/workflows/release.yml	Downgraded two Namespace runner profiles from -large variants to standard profiles for the Linux and macOS tarball build jobs; no other logic changed.

_{Reviews (1): Last reviewed commit: "chore(ci): use non-large runners for rel..." | Re-trigger Greptile}

[github-actions]

Hyperfine Performance

mise x -- echo

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
mise-2026.5.6 x -- echo	21.6 ± 1.4	18.6	26.8	1.00
mise x -- echo	23.0 ± 1.5	20.0	39.4	1.06 ± 0.10

mise env

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
mise-2026.5.6 env	21.7 ± 1.9	18.9	30.7	1.05 ± 0.12
mise env	20.7 ± 1.6	17.7	31.7	1.00

mise hook-env

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
mise-2026.5.6 hook-env	20.5 ± 1.0	18.4	25.5	1.01 ± 0.07
mise hook-env	20.2 ± 1.0	18.3	25.6	1.00

mise ls

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
mise-2026.5.6 ls	20.9 ± 3.9	14.4	27.9	1.28 ± 0.24
mise ls	16.3 ± 0.8	14.7	20.5	1.00
✅ Performance improvement for ls is 28%

xtasks/test/perf

Command	mise-2026.5.6	mise	Variance
install (cached)	126ms	128ms	-1%
ls (cached)	58ms	58ms	+0%
bin-paths (cached)	62ms	64ms	-3%
task-ls (cached)	545ms	552ms	-1%