[MBL-1162] Fix login flow styling

[ifosli]

📲 What

Update login flow classes to use self.traitCollection.userInterfaceIdiom whenever we want to find out if we should use iPad styling or not. The choice of styling here came from me sending Alison a few screenshots and getting her opinion.

Note: I've also updated tests to significantly reduce the number of snapshots we need to deal with; instead of checking all possible combinations, we create a minimal set of combinations where each option is included.

Note: This pr does NOT get rid of every instance of subview.traitCollection; there are too many :(

🤔 Why

Styling in the login flow has gotten weird in iOS 17 for a few reasons. First, traitCollections don't always propagate correctly from viewcontrollers to subviews (though they sometimes seem to?). This means that we should use self.traitCollection instead, particularly for tests. Second, the login flow is presented using a modal overlay, which means that the view itself has a compact horizontal size class. This means that we should check that the idiom is .pad instead of checking the size classes, since we do still want different styling on iPad.

Note: I kept the two factor verification screen in terms of size classes instead, since it got so compressed that the english sentence got truncated at the default font otherwise.

👀 See

JIRA

Before 🐛	After 🦋

♿️ Accessibility

Not perfect, but also not worse. Log in button is hidden (scroll down to see it) for larger font sizes.

✅ Acceptance criteria

• Tests are representative of app
• Login flow looks decent

[scottkicks]

Looks good. I'll admit that I don't fully understand how those orthogonal methods work exactly yet though

[ifosli]

Looks good. I'll admit that I don't fully understand how those orthogonal methods work exactly yet though

Let's say we want to test [a, b] traits from one group (ex languages) and [x, y, z] traits from another group (ex devices). Our current combos will test every combination; ie [(a, x), (a, y), (a, z), (b, x), (b, y), (b, z)] for a total of six screenshots. The orthogonal combos ensures each trait is represented at least once, by combining them linearly; ie [(a,x), (b, y), (a,z)] for three screenshots. (It tests a twice because there aren't as many traits from the first group, so it starts looping through them again).

Most of our tests are run either 10 or 15 times (5 languages and 2 or 3 devices) - with this change, we'll only run them 5 times. And, if we have multiple similar tests within one class, we don't even need to test each one of those 5 times; once each is probably enough, as long as we vary the device and language.

[scottkicks]

@ifosli got it! thanks for explaining!

[ifosli]

Updating the spacing again with Alison's okay, including "after" screenshots. Spacing is final now!