[RFC] Add --mode test to quasar-cli

[nothingismagick]

This RFC is about integrating a new mode into quasar-cli for scaffolding test runners and actually running tests with the quasar test command.

Goals

The goal of the new mode is to make it dead-simple to integrate the testing framework of the user's choice into their new or existing quasar-framework project. The most time-consuming part of rigging test runners is their configuration, and that is where we hope to relieve the development team.

The testing approach needs to be simple enough for a beginner to understand and use, yet powerful enough to enable requirements-based testing, node api hooks and continuous integration without convoluted and potentially breaking changes to a working quasar project or its babel / webpack configuration. We should offer a "recommended" approach with sensible defaults, but enable professional developers the opportunity to pick and choose test-runners that they are more comfortable with.

To that end, we need to make as many test runners available as possible, while considering the fact that different projects and teams have different needs. At the very least, it must be possible using some combination of test runners to achieve passing unit tests that dive into the <template>, the <script> and produce accurate coverage approaching ~99% without actively ignoring parts of the pre-babel'd code base. Further, e2e / integration tests must enable the developer to gain insight into the built project, whether that is on a quasar dev server, a cordova emulator or even an electron package. Finally, in the interests of code quality, it is desirable to inspect both the final results with e.g. Lighthouse and code bloat via profiling with e.g. the import-cost node package.

Requirements

The "test mode" must:

• be optional
• available for integration during quasar init
• available for integration at any point after quasar init
• be non-destructively removable
• automatically added if the user runs quasar dev -m test
• configurable according to user preferences for test runner frameworks
• reconfigurable if the currently selected runner/s should be changed
• update-able from userland (yarn update) as well as with new versions of quasar-cli
• chainable in order to have any combination of unit, e2e and quality run
• be delivered with at least one passing test that is integrated with both Vue and Quasar

There must be "test runners" available for:

• spa / pwa
• electron
• cordova

The "test mode" should:

• run within a webpack context, i.e. quasar dev test (not possible e.g. with lighthouse)
• run outside of webpack, i.e. quasar test
• run as npm/yarn commands, i.e. yarn test:e2e
• enable parallel execution of e.g. quasar dev in one console and ava --watch in another
• be scriptable and extendable from userland
• make snapshots available where possible
•  centralise all test configs and settings in the project's package.json

The "documentation" will:

• describe testing in general
• have a quick start section describing the folder structure and available commands
• have a basic info page for each of unit, e2e and quality tests
• have a detailed page for each test runner, including its configs, settings and files
• have a troubleshooting page
• be available through the general documentation, but hosted in the monorepo

Link to new monorepo: https://github.com/quasarframework/quasar-test

Matching Quasar Starter Kit RFC regarding quasar init
quasarframework/quasar-starter-kit/issues/43

[nothingismagick]

I am closing this issue because the time for comments is now over.