assuming not a purely financial gain (this would be about control, ergo politics)
Assuming that the value of coins will drop if the attack persists for long enough, then a majority attacker will incur a continuous and ongoing cost to attack (censor) the network. To maintain an attack indefinitely, the attacker would need to keep burning money (in the form of electricity costs), forever. This is part of the economic game theory that helps keep miners 'honest': it is presumedly very expensive to misbehave!
If they ever stop attacking, the network could resume operation, though obviously a prolonged attack could damage faith in the network's operational guarantees (and this may be the goal of a politically motivated attacker).
If money is truly no concern, then the economic incentives that govern the game theory of bitcoin mining will of no concern, so then it just becomes a rather pragmatic consideration of "can any one jurisdiction actually acquire a majority of mining hardware, and energy"?
what could prevent such an attack (especially if pin-point targeted violence is an option)
Since you mentioned targeted violence, its worth pointing out that destroying the attacker's mining equipment would stop the attack. Likewise, destroying the honest miner's equipment will benefit an attacker. If a nation state decided to attack bitcoin, then a coordinated takedown of competing hashpower centres may well be a part of their strategy.
On manufacturing: given the competitive nature of bitcoin mining, and the relatively few high-tech computer chip foundries that currently exist, it might be expected that the world would notice if any one nation-state/etc decided to buy out the entire production capacity of new ASIC miners. Which is to say: it would be tough to organize such a plan covertly, so perhaps proactive prevention is more attainable than reactive prevention.
and how can Bitcoin protect itself (Andreas Antonopoulos says that the attacker would be kicked out... how would that happen in practice?), assuming there are defense mechanisms possible?
I think Andreas is technically incorrect here: there is no way to know which blocks are created by an attacker, so there is no way to selectively ban some specific entity's blocks. Even if you somehow figure out that a certain block was created by an attacker, what guarantee will you have of also discerning the attacker's next block? What happens if you have false positives? Etc.
He mentions that bitcoin users could "rework the protocol around them", but switching to a different hashing algorithm would effectively mean creating a new network, so it's easily arguable that the bitcoin network itself would die in this situation. Further, such a situation would require a messy social consensus to enact, which is exactly the sort of coordination mechanism the bitcoin network seeks to avoid. For example, what if multiple different groups decide to each create their own 'new, gov-proof bitcoin network'? It is extremely unclear how such a situation would play out.