Add prefer-true-attribute-shorthand rule

[g-plane]

Closes #775
Closes #1781

[ota-meshi]

LGTM! Thank you!

[FloEdelmann]

Thanks for implementing this! However, I have a few remarks.

In #775 (comment), it was suggested to make it configurable whether to always or never use shorthand props. I think that's a good idea, given some people may prefer to never use shorthand props because of their quirks or just by taste.

The never case could also be auto-fixable, while the suggestion you implemented would be fine for the always case.

[FloEdelmann]

Nice, this looks really good!

Only one problem I just noticed: The never option will also report native HTML shorthand attributes: <input checked> should become <input checked="checked">, not <input :checked="true">. Or is this actually allowed?

If it's allowed, please also add this test case to the "allowed" list to document this behavior.

If it isn't, we need a little more logic: Since one can use native HTML shorthand attributes like hidden also on components (e.g. <MyComponent hidden>), I think we also have to use suggestions instead of an autofix there: One for changing it into a longhand Vue prop, one for changing it into a longhand HTML attribute.

We might be a bit more clever about this and actually do autofix it in one case though: The shorthand attribute is on a known HTML element. Then we know that it should be autofixed to a longhand HTML attribute. But I think this is a nice-to-have improvement; only having suggestions in all cases would be fine for me.

[FloEdelmann]

See also https://v3-migration.vuejs.org/breaking-changes/attribute-coercion.html

[g-plane]

What about skipping native HTML elements, and just check components?

[FloEdelmann]

That'd be fine for me, but still not catch the <MyComponent hidden> case. hidden might be a prop or an HTML attribute.

[g-plane]

We can turn auto fix into suggestions, as you said.

[FloEdelmann]

Nice, thanks for the improvements and the test cases! Please run npm run update one more time to update the docs (there is no autofix anymore). But this is not merge-blocking; we can do it afterwards.