I do some of that with a content rule:
Sender-blacklist: if (mail-from-dictionary-match("blocked-senders", 1)) { drop(); }
In the blocked-senders dictionary you would then probably need a regex to avoid lots of false positives, perhaps \.sex$ for one of your examples; fortunately there isn't that many of them.
My example uses a drop(), and if experimenting you'd probably want to start with something less permanent, say a quarantine action. There will also be more efficient ways of coding this if you're into message filters.
Is there a list of hopeless TLDs? Not that I know of, and with a few notorious exceptions the real reason the industry keeps dreaming up these new ones is to sell them to the nuisance brigade. If considering reversing my logic with a dictionary of good TLDs skipping a subsequent unconditional drop action, beware that there may be more valid TLDs than you expect.
Looking at alternatives, do you have Envelope Sender Verification turned on?
