Improve CircleCI workflow

[kefranabg]

Feature request

I've noticed that some improvements could be added to the CI process:

• Split the single main process into multiple jobs.
  Why ?

  • From a github PR, it would be easier to understand what's wrong when CI fails.
    We'd have this:
    

    Instead of this:
    

  • Ci process would be faster because we could parallelize some jobs.

• Saving cache should rely on yarn.lock checksum instead of package.json.

• We should use --frozen-lockfile flag when installing deps with yarn (on CI).

• Current docker image is circleci/node:10.15.2 but I think we should use circleci/node:latest as I don't see any compatibility note telling that we shouldn't use latest versions of node 🙂

• Current lint command fix errors (so it never fails). The CI process should never change source code. CI lint command has to be replaced by a lint:check that just check if code is correctly formatted and exit with error code if necessary.

Are you willing to work on this yourself?

Yes 👍I'm on vacation for the moment, but I'll do it as soon as I get back 🏖