♻️ simplify dead code and test boundaries

[Robdel12]

Why

This codebase had gotten pretty bloated. We found production modules that nothing imported, CLI validators that always returned no errors, exports used only by tests, and a pile of tests that mostly proved mocks or asserted nothing (assert.ok(true) included). That is a lot of maintenance cost without much confidence.

What changed

• Removed dead CLI, config, project, TDD, reporter, and client code.
• Removed no-op validation wiring and internal-only test seams.
• Reworked command coverage around observable behavior: pure tests for pure logic, real CLI and local HTTP workflows at command boundaries, and browser workflows for the reporter.
• Removed tests that reached into private pool metadata or mocked internal modules.
• Preserved and contract-tested the public Ember browserMappings export after the final audit caught that @vizzly-testing/ember/testem exposes it.
• Made the unused events-router JSON helper private and removed the remaining internal TDD dependency-wiring test.

The goal here is not a better coverage number. The goal is tests that tell us when a user workflow broke. We kept the tests at the highest useful level and left the old mock-heavy suites behind.

Evidence

The root suite and every client suite pass. Storybook passes 172 tests, static-site passes 205, Ember passes 45, Vitest passes 23, and the reporter workflows pass 35 with 5 environment-appropriate skips. The build, published type checks, lint, formatting, and diff checks are clean. A final repository sweep found no remaining assert.ok(true) tests, deleted-module references, or dangling references to the removed code.

Follow-up

Real baseline-download success coverage and a few broader preview/upload failure scenarios would be useful in a separate boundary-focused test pass. I did not bring back the old broad mock suites just to fill those gaps.