feat: integrate TinyTorch into MLSysBook repository

TinyTorch educational deep learning framework now lives at tinytorch/

Structure:
- tinytorch/src/         - Source modules (single source of truth)
- tinytorch/tito/        - CLI tool
- tinytorch/tests/       - Test suite
- tinytorch/site/        - Jupyter Book website
- tinytorch/milestones/  - Historical ML implementations
- tinytorch/datasets/    - Educational datasets (tinydigits, tinytalks)
- tinytorch/assignments/ - NBGrader assignments
- tinytorch/instructor/  - Teaching materials

Workflows (with tinytorch- prefix):
- tinytorch-ci.yml           - CI/CD pipeline
- tinytorch-publish-dev.yml  - Dev site deployment
- tinytorch-publish-live.yml - Live site deployment
- tinytorch-build-pdf.yml    - PDF generation
- tinytorch-release-check.yml - Release validation

Repository Variables added:
- TINYTORCH_ROOT  = tinytorch
- TINYTORCH_SRC   = tinytorch/src
- TINYTORCH_SITE  = tinytorch/site
- TINYTORCH_TESTS = tinytorch/tests

All workflows use \${{ vars.TINYTORCH_* }} for path configuration.

Note: tinytorch/site/_static/favicon.svg kept as SVG (valid for favicons)

tinytorch/tests/cli/README.md

@@ -0,0 +1,145 @@
+# CLI Test Suite
+
+Comprehensive test suite for the TinyTorch CLI (`tito`). Ensures all commands are properly registered, documented, and executable.
+
+## Test Files
+
+### 1. `test_cli_registry.py` - Command Registration Tests
+Validates the command registry in [tito/main.py](../../tito/main.py#L74-L99):
+
+- **TestCLIRegistry**: Core registry validation
+  - All commands inherit from `BaseCommand`
+  - All commands have descriptions (>10 chars)
+  - All commands implement `execute()` and `add_arguments()`
+  - Parser creation succeeds
+  - No duplicate command names
+  - Help text accessible for all commands
+
+- **TestCommandFiles**: File system consistency
+  - All registered commands have corresponding files
+  - No orphaned command files (files without registration)
+  - Detects commands that need cleanup
+
+- **TestEpilogDocumentation**: Help text consistency
+  - Epilog mentions all registered command groups
+  - No outdated references
+
+### 2. `test_cli_execution.py` - Command Execution Tests
+Smoke tests for actual command execution:
+
+- **TestCommandExecution**: Basic execution
+  - Bare `tito` command shows welcome screen
+  - `tito -h` shows help
+  - `tito --version` works
+  - All 18 commands can show help (`tito <cmd> -h`)
+  - All subcommands can show help
+
+- **TestCommandGrouping**: Discoverability
+  - Student-facing commands visible in welcome screen
+  - Developer commands documented in help
+
+- **TestErrorMessages**: Error handling
+  - Invalid commands show helpful errors
+  - Missing subcommands show help
+
+### 3. `test_cli_help_consistency.py` - Help Text Quality
+Ensures help text is consistent and complete:
+
+- **TestHelpConsistency**: Formatting consistency
+  - All command helps mention 'tito'
+  - Bare `tito` vs `tito -h` serve different purposes
+  - No references to removed commands
+
+- **TestWelcomeScreen**: Welcome screen quality
+  - Shows Quick Start section
+  - Organizes commands into groups
+  - Includes example commands
+
+- **TestCommandDocumentation**: Documentation completeness
+  - All registered commands documented
+  - Checkpoint command specifically validated
+
+## Running the Tests
+
+Run all CLI tests:
+```bash
+pytest tests/cli/ -v
+```
+
+Run specific test file:
+```bash
+pytest tests/cli/test_cli_registry.py -v
+```
+
+Run with output:
+```bash
+pytest tests/cli/ -v -s
+```
+
+## What These Tests Catch
+
+### ✅ Prevents
+
+1. **Orphaned Commands**: Command files that aren't registered
+2. **Missing Documentation**: Commands without descriptions
+3. **Broken Help**: Commands that crash when showing help
+4. **Inconsistent UX**: Different help formats across commands
+5. **Stale References**: Help text mentioning removed commands
+
+### ✅ Validates
+
+1. **Registration**: All commands in `TinyTorchCLI.commands` dict
+2. **Implementation**: All commands inherit from `BaseCommand`
+3. **Documentation**: All commands have clear descriptions
+4. **Execution**: All commands can run without crashing
+5. **Discoverability**: Key commands visible to users
+
+## Test Coverage
+
+**52 tests** covering:
+- 18 registered commands
+- 8 subcommand groups
+- Registry validation
+- Execution smoke tests
+- Help text consistency
+- Welcome screen UX
+
+## How Modern CLIs Handle Testing
+
+These tests follow industry best practices from tools like:
+
+1. **Click** (Python): Command registration validation
+2. **Git**: Comprehensive help text testing
+3. **Docker**: Smoke tests for each command
+4. **kubectl**: Subcommand hierarchy validation
+
+Key patterns used:
+- **Registry-based validation**: Single source of truth in code
+- **Smoke tests**: Don't test functionality, just "does it run?"
+- **Help text parsing**: Ensure documentation stays current
+- **Snapshot testing**: Compare outputs (could be added later)
+
+## Maintenance
+
+When adding a new command:
+
+1. Add to `TinyTorchCLI.commands` dict in [tito/main.py](../../tito/main.py#L74-L99)
+2. Create command file in `tito/commands/`
+3. Add to epilog if it's a major command group
+4. Tests will automatically validate it!
+
+When removing a command:
+
+1. Remove from `TinyTorchCLI.commands` dict
+2. Delete command file OR add to `known_internal` in tests
+3. Update epilog/welcome screen
+
+## Future Enhancements
+
+Consider adding:
+
+- **Snapshot tests**: Save known-good help output, compare changes
+- **Integration tests**: Test actual command workflows
+- **Performance tests**: Ensure CLI startup is fast
+- **Completion tests**: Validate shell completion scripts
+- **Config tests**: Test config file parsing and validation