Remove stale target tests

[tybug]

I'm taking a look at improving cover test times. test_targeting_with_following_empty takes 3-5 seconds to run - target(len(lst)) is dangerous, as my observation is hypothesis occasionally tries doubling the length of the list, which can quickly get out of hand and cause slow generations when unbounded.

I was going to simply improve the runtime. However, I don't think the test (and the related test_targeting_with_many_empty) is needed anymore. They were introduced in #2137, back when there was special-case handling for empty vs non-empty examples in Optimiser. This logic was changed in #2289, and I don't think there's a distinction between empty and non-empty anymore, but please correct me if there is.

[Zac-HD]

I've cherry-picked this into #3769 with some other minor fixes, and queued it to merge and release 🎉

Thanks @tybug!