1) The differences are subtle (and may vanish altogether in the next beta, we're still going back and forth on that) - generally you'd want to use Chinese Example for sentences. You should be able to pick a card type in a menu at the top of the screen when you create a new custom card.
2) They're hyperlinks, yes. I wouldn't recommend trying to mess around with them directly yet, but basically, Cloze Links link to specific sentences (can be other cards or dictionary ones) that you want to use for cloze testing (we'll pick other sentences by default if you don't have any but if you do pick some then we'll use those) - you can modify that via the Organize definition screen if you've turned on the option for a + for each example; 'reader example' links are context sentences created from the document reader (settings / reader), and 'variant info' is something we only use internally to map onto an equivalent field in Pleco-created dictionaries.
3) That should also be possible in 'entry body', is it?
4) * (actually ⁕) is a token skip character; you can insert them in the character or pinyin field to make the lengths of those match up for the purposes of tone coloring / ruby / etc. // (actually ᜶) is for optional patterns, the same as in Pleco dictionaries - a space in a word where you can insert other words; our document reader uses these so that e.g. with 犯//愁 you could tap on the 犯 or the 愁 in 你犯什么愁 and it would show you the entry for 犯愁 along with the one for whichever of 犯/愁 you'd tapped on.
5) Could you be more specific about these 'visual hints'? In general the preferred way to do clozes in 4.0 is to make a card for the word you want to test and then link to a cloze from that Cloze Links field (or not bother and rely on us to pick a sentence for you), since that way we can track statistics for the word consistently (even if you're not always testing it in a cloze) and can easily switch between multi-choice / stroke order / etc tests for the same card because we know all of the details about the underlying word.
There isn't really a great way to have a custom card to do what you're describing with a default test type - the best option I can think of would be to put the sentence + visual hints in the 'entry body' field, headword in the appropriate Chinese headword field, and the definition in the 'notes' field; I *think* in the current version that if you did a test 'showing' definition' it would show just the 'entry body' field at first and then show the others (including notes) when the card is revealed, though I'd suggest testing this with one or two cards before you build any more around it.
With a custom test type you can show whatever combination of fields you like whenever you like, but those are *extremely* hairy to construct right now; for the next beta we've added a streamlined version where you don't have to muck around with 'presentations' and can simply say that you want these fields in this order with this formatting in the front/back of the card, so if neither of the above options works, you could wait for that beta and there should be a better solution then.
The main question is whether a feature using it for something not camera related would pass App Review, but it wouldn't hurt to try. (we could always have it default to opening OCR - a usage they're unlikely to have a problem with - but then let you change it to something else in settings)