ci(l1): optimize CI times with toolchain caching

[ilitteri]

Motivation

CI runs are taking longer than necessary due to repeated downloads of zkVM toolchains and suboptimal workflow configuration. This PR implements several optimizations to reduce CI time.

Description

Toolchain Caching

• SP1: Cache ~/.sp1 directory with key sp1-toolchain-5.0.8
• RISC0: Cache ~/.risc0 directory with key risc0-toolchain-3.0.3
• ZisK: Cache ~/.zisk directory with key zisk-toolchain-0.14.0
  • Also fixed the install script to use versioned URL (v0.14.0) instead of latest

When updating toolchain versions, update both the install command AND the cache key.

Workflow Optimizations

• Remove fetch-depth: 0 from checkout (shallow clone is sufficient for linting)
• Change trigger from push to all branches → pull_request + push to main only
  • Prevents duplicate CI runs (PR creation + branch push were both triggering)
• Combine rustup component add rustfmt and clippy into single command

Free Disk Space

Expanded cleanup to include:

• CodeQL bundles
• Swift
• PowerShell
• Miniconda

Expected Impact

Optimization	Expected Savings
Toolchain caching	1-3 min per job (on cache hit)
Shallow checkout	5-10s per job
Single rustup command	~2s per job
Fewer duplicate runs	50% fewer total runs

How to Test

1. Merge this PR
2. Open a new PR to trigger CI
3. First run will populate caches (no speedup)
4. Subsequent runs should show cache hits and faster times

Checklist

• SP1 cache key includes version
• RISC0 cache key includes version
• ZisK cache key includes version
• ZisK install uses versioned script
• Trigger changed to PR + main
• Free disk expanded

[Copilot]

Pull request overview

This PR optimizes CI run times through toolchain caching, workflow trigger improvements, and enhanced disk space cleanup. The goal is to reduce redundant downloads and eliminate duplicate CI runs.

Changes:

• Added caching for SP1, RISC0, and ZisK toolchains with versioned cache keys
• Changed workflow triggers from all branch pushes to pull requests and main branch pushes only to prevent duplicate runs
• Consolidated multiple free disk space cleanup commands and added cleanup for CodeQL, Swift, PowerShell, and Miniconda
• Combined rustup component installation commands and removed unnecessary fetch-depth parameter

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
.github/workflows/pr-main.yaml	Updated workflow triggers to reduce duplicate runs, removed fetch-depth, and consolidated rustup commands
.github/actions/install-sp1/action.yml	Added caching for SP1 toolchain with version 5.0.8
.github/actions/install-risc0/action.yml	Added caching for RISC0 toolchain with version 3.0.3
.github/actions/install-zisk/action.yml	Added caching for ZisK toolchain with version 0.14.0, fixed install script to use versioned URL, separated PATH update into dedicated step
.github/actions/free-disk/action.yml	Consolidated cleanup commands into single step and added additional cleanup targets

Comments suppressed due to low confidence (2)

.github/actions/install-sp1/action.yml:19

• The SP1 binaries need to be added to PATH regardless of whether the cache was hit. When the cache is restored, the installation step is skipped, but the PATH is not updated. This will cause failures when trying to use SP1 tools. Add a separate step after the installation that always runs to add the SP1 bin directory to GITHUB_PATH, similar to how it's done in the ZisK installation action (line 28-30 in install-zisk/action.yml).

    - name: Install SP1
      if: steps.sp1-cache.outputs.cache-hit != 'true'
      shell: bash
      run: |
        curl -L https://sp1up.succinct.xyz | bash
        ~/.sp1/bin/sp1up --version 5.0.8

.github/actions/install-risc0/action.yml:23

• The RISC0 binaries need to be added to PATH regardless of whether the cache was hit. When the cache is restored, the installation step is skipped, but the PATH is not updated. This will cause failures when trying to use RISC0 tools. Add a separate step after the installation that always runs to add the RISC0 bin directory to GITHUB_PATH, similar to how it's done in the ZisK installation action (line 28-30 in install-zisk/action.yml).

    - name: Install RISC0
      if: steps.risc0-cache.outputs.cache-hit != 'true'
      shell: bash
      run: |
        curl -L https://risczero.com/install | bash
        ~/.risc0/bin/rzup install cargo-risczero 3.0.3
        ~/.risc0/bin/rzup install risc0-groth16
        ~/.risc0/bin/rzup install rust
        ~/.risc0/bin/rzup install cpp
        ~/.risc0/bin/rzup install r0vm 3.0.3

---

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

[Copilot]

The system dependencies are installed unconditionally every time, even when the ZisK toolchain is restored from cache. While this ensures dependencies are present, it adds unnecessary overhead on cache hits. Consider adding a conditional check or caching the apt packages separately if they are solely for ZisK and not used elsewhere in the workflow.