Add comprehensive pytest suite (134 tests) + fix 3 bugs it found

[SomethingNew71]

Summary

Adds an automated pytest test suite covering the entire WireViz API surface. 134 tests across 9 files, completing in ~5 seconds. Every upstream PR we ported and every gemini review fix gets a pinned regression test, so future refactors can't silently undo any of the work that landed in PRs #1-#10.

The suite already paid for itself by surfacing three real bugs, all fixed in this same PR.

Test files

File	Tests	Coverage
tests/test_smoke.py	10	Every output format renders + has the right magic bytes
tests/test_parse.py	20	wireviz.parse() API: input shapes, return_types, embed_yaml, source_path auto-fill
tests/test_cli.py	19	Every CLI flag via Click's CliRunner — stdin/stdout, .png input, --no-embed-yaml, -t, -p, error paths
tests/test_harness.py	11	Harness public methods + _render dict shape
tests/test_dataclasses.py	19	Connector / Cable / Tweak / Options / Image coercion + validation
tests/test_colors.py	12	DIN/IEC/T568/TEL color schemes, hex parsing, padding heuristic
tests/test_bom.py	8	BOM aggregation, ignore_in_bom, additional_bom_items, bundles, part numbers
tests/test_regressions.py	29	One test per upstream-PR port + every gemini review fix
tests/test_round_trip.py	6	PNG embed/extract round-trip, stdin→stdout, dict-input no-mutation

Bugs found and fixed

1. Python 3.13+ re.sub positional-count deprecation

wv_html.py:72 was calling re.sub(pattern, repl, string, 1) — passing count positionally. As of Python 3.13 that emits a DeprecationWarning; pytest's filterwarnings = "error" config turns it into a test failure. Fixed by switching to count=1 keyword form. Would have broken CI on Python 3.13 once that DeprecationWarning becomes a real error.

2. CLI "File does not exist" raised generic Exception

wv_cli.py had two raise Exception(f"File does not exist:\n{file}") sites — one for the input file, one for --prepend. Both produced a Python stack trace for users instead of the clean Click error footer. Promoted both to click.UsageError, matching the pattern we'd already applied to the unknown-format check earlier.

3. parse(dict) silently mutating caller's dict

This is the most consequential one for the upcoming GUI work. When parse() was called with a pre-parsed dict, it rewrote the connections section in-place during processing — [[{X1: [1, 2]}, {W1: [1, 2]}]] became [[{X1: 1}, {X1: 2}], [{W1: 1}, {W1: 2}]] on the caller's reference. The GUI will likely keep an in-memory model of the harness and call parse() repeatedly on it; without this fix every render would corrupt the model.

Fixed by copy.deepcopy(inp) in _get_yaml_data_and_path. Pinned with tests/test_round_trip.py::test_dict_input_not_mutated.

Also: cherry-pick of e4f485e (PR #10 review fixes)

The PR #10 review-fix commit was on port-321-stdin-stdout but never reached master because PR #10 merged before that commit was pushed. Cherry-picked here so the test suite can validate:

• Data URIs no longer have a leading space (RFC 2397)
• -f X (unknown format) raises click.UsageError, not generic Exception
• Stdin mode initializes image_paths = {Path.cwd()} so relative image: src: paths can resolve
• Tweak override conflicts report both the existing and new value
• source_path auto-fills from a Path input (matching the docstring)

Infrastructure

• pyproject.toml gains [tool.pytest.ini_options] with testpaths = ["tests"] and filterwarnings = ["error", "ignore::SyntaxWarning"] so future deprecation drift gets caught.
• tests/conftest.py centralizes fixtures pointing at small YAMLs in tests/fixtures/ — deliberately distinct from the gallery YAMLs in examples/ so tests don't break when the gallery shifts.
• New .github/workflows/tests.yml runs pytest across Python 3.7-3.12 (same matrix as the existing Create Examples workflow).
• CLAUDE.md updated to document the test suite layout and how to run it.

Verification

```
$ pytest -v
============================= 134 passed in 5.08s ==============================
```

Both pytest and build_examples.py pass green. The two are complementary — pytest validates the API/CLI/data-model surface, the example sweep validates rendered visual output across the gallery.

Test plan

• CI's new Tests workflow passes across Python 3.7-3.12
• CI's existing Create Examples workflow still passes (no rendered output changes)
• Spot-check that adding a new feature is now natural to TDD via the regression file pattern

🤖 Generated with Claude Code

[gemini-code-assist]

Code Review

This pull request introduces a comprehensive automated test suite using pytest, covering the CLI, API, BOM aggregation, and various internal logic components. It also includes several bug fixes and improvements, such as preventing the mutation of input dictionaries through deep copying, refining data URI formatting by removing illegal whitespace, and enhancing CLI error reporting with specific usage errors. Documentation and CI configurations have been updated to reflect the transition to automated testing. Feedback is provided regarding the placement of an import statement in src/wireviz/wireviz.py to better align with PEP 8 standards.

[gemini-code-assist]

The import copy statement is placed inside the _get_yaml_data_and_path function. According to PEP 8, imports should be at the top of the file. Moving this import to the top-level section improves code clarity and ensures consistency with standard Python practices.