feat(cli): unbundle 3rd-party CLIs

[louis-bompart]

Proposed changes

Currently, a large part of the footprint of the CLI is caused by @vue/cli or @angular/cli, or create-react-app.
Unbundling them solves the issue and also ensures that the user will have a smooth ride moving forward when using either of them.

Removed some unused branching in the preconditions factories and add support for auto-fix in precondition

Breaking changes

This is arguable. This could be considered a breaking change on ui:create:angular because we now have a new user interaction branching. However, I do not consider CI/CD usage of this command to be a use case we want to support. Therefore, I'll consider the breaking aspect on iff the input of the command changed (no), or if the output of the command changed (no)

Testing

• Unit Tests: 💯 unit-tested
• Functional Tests: If the current tests pass (In the CI I trust 🎲), then it's covered, given that @angular/cli is the tricky bit with the global install, and is not installed in the setup phase (which mean it'll therefore be installed by the preconditions and the Terminal manipulation.
• Manual Tests: Ran the three commands

---

CDX-478

[louis-bompart]

• revert(ui): add better message when user doesn't have permission #436

[y-lakhdar]

[olamothe]

💯 Nice job

[louis-bompart]

#434 (comment)
#434 (comment)

This is a funny story. I'll start with the short version: Windows. The long one now:

Node.JS rely on libuv/libuv for IO operations (which is fine!). However, libuv/libuv also does that:
https://github.com/libuv/libuv/blob/v1.x/include/uv/errno.h#L245-L249

What this does, is 'if ENOENT is defined, then I set UV__ENOENT to its value, else, I set it to -4058'
POSIX system has an ENOENT defined at a known and easy value: 2 (KISS).
Windows, you guessed it, has not.

So, when a file is missing on Windows, we got a -4058 for errorCode, and 2 for POSIX.
But the fun does not stop there, remember, 'If ENOENT is defined, then I set UV__ENOENT to its value'

Well, Windows user can define ENOENT (or so I read) if they so desire . Which that the errno is worth nothing and can't be relied upon.

I'll double-check something tho on that given that 'code' is consistent it seems:

• Check if we could use code