ci: enable sccache GHA backend and optimize rust-cache keys

[staging-devin-ai-integration]

Summary

sccache has been running with a 0% cache hit rate on all hosted runner jobs because the SCCACHE_GHA_ENABLED environment variable was never set. The mozilla-actions/sccache-action@v0.0.9 installs sccache and exposes the GHA cache service variables but — unlike earlier versions — does not set SCCACHE_GHA_ENABLED=true. Without it, sccache falls back to local disk (/home/runner/.cache/sccache), which is empty on each ephemeral runner.

Additionally, release.yml and marketplace-build.yml installed sccache but never set RUSTC_WRAPPER=sccache, so it was never used as the compiler wrapper.

This PR fixes these issues and optimizes the Swatinem/rust-cache configuration:

1. Enable sccache GHA backend — adds SCCACHE_GHA_ENABLED: "true" to all 5 Rust workflow env blocks (skit.yml, e2e.yml, plugins.yml, release.yml, marketplace-build.yml).

2. Add RUSTC_WRAPPER: sccache to release.yml and marketplace-build.yml where it was missing.

3. Share rust-cache via shared-key — groups 7 independent job caches into 3 shared namespaces:

   • skit → lint, test, build (ubuntu-22.04)
   • skit-e2e → pipeline-validation, playwright-e2e (ubuntu-22.04)
   • skit-gpu → test-gpu, pipeline-validation-gpu (self-hosted)

4. Save rust-cache only from main — adds save-if: ${{ github.ref == 'refs/heads/main' }} on hosted runner jobs so PR pushes don't churn the cache pool. sccache (now with GHA backend) handles the delta for changed crates.

5. Enable cache-on-failure: true on all rust-cache entries so partially-completed builds still contribute.

Evidence from a recent Skit / Test run showing the problem:

Cache hits                            0
Cache misses (Rust)                  77
Cache hits rate                    0.00 %
Cache location             Local disk: "/home/runner/.cache/sccache"

The self-hosted GPU runner (persistent disk) confirms sccache works when it has a warm cache:

Cache hits (Rust)                    11
Cache hits rate (Rust)           100.00 %

Review & Testing Checklist for Human

• Verify sccache stats in the first CI run after merge — the Skit / Test and Skit / Lint post-step should show Cache location: GitHub Actions Cache instead of Local disk and a non-zero hit rate on subsequent runs
• Check https://github.com/streamer45/streamkit/actions/caches after a few main-branch runs to confirm the number of v0-rust-* entries has decreased (3 shared-key groups instead of 7 per-job groups)
• Confirm PR-branch CI runs still restore caches correctly (they should restore from main's cache but not save new entries)

Notes

• The self-hosted GPU jobs (skit-gpu shared-key) do not have save-if since the GPU runner's persistent local sccache disk already provides excellent hit rates, and always-save ensures the GHA rust-cache stays warm as a fallback.
• sccache inherently cannot cache cargo clippy invocations (~85% of lint calls are "multiple input files") or proc-macro/cdylib crate types (~50% of test/build). rust-cache continues to handle these via target/ snapshot restore.
• The plugins.yml rust-cache already had cache-on-failure: true and a custom workspaces config — only SCCACHE_GHA_ENABLED was added there.

Link to Devin session: https://staging.itsdev.in/sessions/cdf841e834eb419aa04168a565311f10
Requested by: @streamer45

[staging-devin-ai-integration]

✅ Reviewed on bb575a2

View review

[staging-devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment and CI monitoring

[staging-devin-ai-integration]

✅ Devin Review: No Issues Found

Devin Review analyzed this PR and found no potential bugs to report.

View in Devin Review to see 4 additional findings.

Debug

Playground