This PR ensures that all DOM nodes ids are unique when multiple instances of a component co-exist on the same page.

It also allows fragment caching of generated HTML (since the unicity does not depend on other components on the page).

Solves #437

A few thoughts following discussion in #437 (numbered, to ease followups):

1. If I understand correctly, dom_id is not only used in tests, but also to attach React components to DOM nodes (you're probably using ReactDOM.render or equivalent)
2. I did not implement the proc version, sorry 😞 (but I'm not sure it would really be useful, tbh)
3. I'd consider my implementation backward-compatible (since it only changes internal behaviour: I consider auto-generated ids to be internal) so I don't think a major bump is needed (at least, if you're using semantic versioning)... OK, actually, I see 1 way of considering generated HTML to be broken: if some users are looking for ids using an hard-coded value (like all the integration tests 😢) or a number-only regexp like /\d+/...
4. Regarding fragment caching, we simply use a unique identifier as id parameter for react_component (in our case an ActiveRecord id, but it could as well be a UUID) and then wraps the call to react_component in a cache(["v1", I18n.locale, some_dependencies...]) block
5. I'm not sure this is how you expected me to make the integration tests pass, let me know if it needs to be changed (maybe we should not explicitly define id for react_component calls that have a display-only purpose?)

I hope this helps.

This change is