i18n - Removes t from props via metamask-connect; put t on context

[danjm]

Fixes #3787

See issue #3787 for details about this. This seems to be working in all places. We should give it another thorough round of QA though.

[kumavis]

e2e error seems really consistent, retried it a few times

[kumavis]

confirmed the QR code view on old-ui does not render (but no error?)
Edit: actually the test does not look for the actual QR code, it just checks that the option is in the menu

[danjm]

• Need to update context after changing language according to the advice here: https://medium.com/@mweststrate/how-to-safely-use-react-context-b7e343eff076

[kumavis]

As an experiment i merged this with #3798
which produces artifacts for test failures https://circleci.com/gh/MetaMask/metamask-extension/9957#artifacts/containers/0

here's the first one:

[danjm]

Need to update context after changing language according to the advice here:

I did further research (e.g. facebook/react#2517) and I think what we currently have is good. We just need to make sure that we don't use shouldComponentUpdate until we switch to react 16.3

shouldComponentUpdate is not used anywhere in our code base yet. Its purpose is to optimize performance, and so is used in often in libraries that are optimizing for that or apps where the combo of component tree size and state update frequency make performance an issue.

I don't think we'll need to worry about it in the near future, and react 16.3 will be production ready in the near future. Meanwhile, local testing now shows consistent and successful updating behaviour. (I think my build was messed up before.)

[kumavis]

somehow CI did not get triggered 🤔

[kumavis]

👏 👏 👏 👏