Migrate from Yarn Classic to pnpm for improved monorepo support

## Summary

Migrate from Yarn Classic 1.22.22 to pnpm for better monorepo support, faster installs, and native workspace linking that eliminates the need for yalc.

## Context

This issue supersedes the "stay with Yarn" decision from #2088, which was made when yalc was still critical to our workflow. Now that #2089 plans to remove yalc (post-monorepo merger), the main argument for staying with Yarn Classic is gone.

### Why Now?

1. **yalc removal is planned (#2089)** - The main blocker for pnpm is being removed
2. **Monorepo maturity** - The monorepo structure is stabilizing (Phase 5 complete, Phase 6-8 in progress)
3. **Clean slate opportunity** - Better to migrate during the restructuring than after

### Current Pain Points with Yarn Classic + yalc

1. Manual `yalc publish` required after changes
2. Not integrated into monorepo workflow
3. Additional tool dependency
4. Requires `.yalc` directory and lock file management
5. Can cause confusion when packages get out of sync
6. Yarn Classic is in maintenance mode (no new features)

## Benefits of pnpm

### 1. Native Workspace Linking (Replaces yalc)
```bash
# In package.json, use workspace protocol
"dependencies": {
  "react-on-rails": "workspace:*"
}

# Changes automatically linked - no manual publish step!
```

### 2. Faster Installs
- Content-addressable storage
- Hard links instead of copies
- Typically 2-3x faster than Yarn Classic

### 3. Better Disk Usage
- Packages shared across projects
- Single global store instead of per-project copies

### 4. Selective Operations
```bash
# Build specific package
pnpm --filter react-on-rails build

# Build all packages recursively
pnpm -r build

# Run tests in one package
pnpm --filter react-on-rails-pro test
```

### 5. Strict Dependencies
- Prevents phantom dependencies
- Catches missing peer deps early
- More reliable builds

### 6. Industry Adoption
- Used by Vue, Vite, Turborepo, and many major monorepos
- Active development with frequent releases
- Growing community and ecosystem support

## Migration Plan

### Prerequisites
- [ ] Complete monorepo merger (Phases 5-8 from #2106)
- [ ] Remove yalc (#2089)

### Phase 1: Preparation (1 hour)
- [ ] Create `pnpm-migration` branch
- [ ] Install pnpm: `npm install -g pnpm`
- [ ] Import lockfile: `pnpm import` (converts yarn.lock to pnpm-lock.yaml)

### Phase 2: Configuration (2 hours)
- [ ] Create `.npmrc` with pnpm settings:
  ```ini
  # Hoist patterns for packages that expect flat node_modules
  shamefully-hoist=false
  strict-peer-dependencies=true
  auto-install-peers=true
  ```
- [ ] Update `package.json`:
  - Set `"packageManager": "pnpm@9.x.x"`
  - Update workspaces config if needed (syntax is compatible)
- [ ] Update workspace package dependencies to use `workspace:*` protocol:
  ```json
  "dependencies": {
    "react-on-rails": "workspace:*"
  }
  ```

### Phase 3: Script Updates (2 hours)
- [ ] Update `package-scripts.yml`:
  - Replace `yarn` with `pnpm`
  - Update workspace commands to use `pnpm --filter`
- [ ] Update `Rakefile` and `rakelib/` scripts
- [ ] Update CI workflows (`.github/workflows/`)
- [ ] Update `bin/` scripts

### Phase 4: CI Configuration (2 hours)
- [ ] Update GitHub Actions to use pnpm:
  ```yaml
  - uses: pnpm/action-setup@v2
    with:
      version: 9
  - name: Install dependencies
    run: pnpm install --frozen-lockfile
  ```
- [ ] Update cache configuration for pnpm store
- [ ] Test all CI workflows pass

### Phase 5: Documentation (1 hour)
- [ ] Update `CLAUDE.md` development commands
- [ ] Update `CONTRIBUTING.md`
- [ ] Update `README.md` installation instructions
- [ ] Update `.conductor/` workspace docs
- [ ] Remove all yalc-related documentation

### Phase 6: Cleanup (30 min)
- [ ] Remove `yarn.lock`
- [ ] Remove `.yalc/` directories
- [ ] Remove `yalc.lock` files
- [ ] Remove yalc from devDependencies
- [ ] Update `.gitignore`

### Phase 7: Validation (1 hour)
- [ ] Run full test suite: `pnpm test`
- [ ] Run linting: `pnpm lint`
- [ ] Test package builds: `pnpm -r build`
- [ ] Test in spec/dummy apps
- [ ] Verify workspace linking works correctly
- [ ] Test clean install from scratch

## Testing Checklist

Before merging, verify:

- [ ] `pnpm install` completes successfully
- [ ] `pnpm -r build` builds all packages
- [ ] Workspace packages are properly linked:
  ```bash
  cd packages/react-on-rails-pro
  ls -la node_modules/react-on-rails  # Should be symlink
  ```
- [ ] Changes propagate automatically (no manual publish needed)
- [ ] All CI workflows pass
- [ ] Example apps work
- [ ] Ruby gem builds correctly
- [ ] NPM packages publish correctly (test with `pnpm pack`)

## Rollback Plan

If issues arise post-migration:

1. Restore `yarn.lock` from git history
2. Remove `pnpm-lock.yaml`
3. Revert script changes
4. Run `yarn install`

The yarn.lock can be preserved in a branch for 30 days before deletion.

## Estimated Effort

| Phase | Time |
|-------|------|
| Preparation | 1 hour |
| Configuration | 2 hours |
| Script Updates | 2 hours |
| CI Configuration | 2 hours |
| Documentation | 1 hour |
| Cleanup | 30 min |
| Validation | 1 hour |
| **Total** | **~10 hours** |

## Risks and Mitigations

### Risk 1: Tool Compatibility
**Risk:** Some tools may not support pnpm
**Mitigation:** pnpm is widely supported; can use `shamefully-hoist=true` as fallback

### Risk 2: CI Cache Invalidation
**Risk:** Initial CI runs will be slower without cache
**Mitigation:** pnpm's subsequent runs are faster; cache will rebuild quickly

### Risk 3: Team Learning Curve
**Risk:** Developers unfamiliar with pnpm
**Mitigation:** Similar commands to yarn; good documentation; minimal mental model shift

### Risk 4: Shakapacker Compatibility
**Risk:** Shakapacker may have yarn-specific assumptions
**Mitigation:** Shakapacker uses npm client detection; test thoroughly in dummy apps

## Success Criteria

- [ ] No yalc usage anywhere in the repo
- [ ] All workspace packages use `workspace:*` protocol
- [ ] Single `pnpm install` sets up entire monorepo
- [ ] Changes in one package immediately visible in dependent packages
- [ ] CI is faster or equivalent to current setup
- [ ] All tests pass
- [ ] Documentation is updated

## Dependencies

- #2089 - Remove yalc (should be completed first or in parallel)
- #2106 - Phase 7-8 Final Monorepo Polish (good timing to include this)

## References

- [pnpm documentation](https://pnpm.io/)
- [pnpm workspaces](https://pnpm.io/workspaces)
- [Migrating from Yarn](https://pnpm.io/cli/import)
- [pnpm GitHub Action](https://github.com/pnpm/action-setup)
- Issue #2088 - Original "stay with Yarn" decision (now superseded)
- Issue #2089 - Remove yalc plan
- `analysis/monorepo/YALC_ALTERNATIVES_ANALYSIS.md` - Detailed analysis

## Labels

enhancement, dependencies

Phase	Time
Preparation	1 hour
Configuration	2 hours
Script Updates	2 hours
CI Configuration	2 hours
Documentation	1 hour
Cleanup	30 min
Validation	1 hour
Total	~10 hours

Uh oh!

Migrate from Yarn Classic to pnpm for improved monorepo support #2120

Description

Summary

Context

Why Now?

Current Pain Points with Yarn Classic + yalc

Benefits of pnpm

1. Native Workspace Linking (Replaces yalc)

2. Faster Installs

3. Better Disk Usage

4. Selective Operations

5. Strict Dependencies

6. Industry Adoption

Migration Plan

Prerequisites

Phase 1: Preparation (1 hour)

Phase 2: Configuration (2 hours)

Phase 3: Script Updates (2 hours)

Phase 4: CI Configuration (2 hours)

Phase 5: Documentation (1 hour)

Phase 6: Cleanup (30 min)

Phase 7: Validation (1 hour)

Testing Checklist

Rollback Plan

Estimated Effort

Risks and Mitigations

Risk 1: Tool Compatibility

Risk 2: CI Cache Invalidation

Risk 3: Team Learning Curve

Risk 4: Shakapacker Compatibility

Success Criteria

Dependencies

References

Labels

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions