You can make BIP first, but Bitcoin Core or other full node software doesn't have to implement such feature.
Of course this is only one possible solution to the "data spam" problem, there may be many others. I'd only thought a bit about this and technically it should be possible, and it was funny that the Wired article proposed that too.
According to recent discussions about Ordinals in the bitcoin-dev mailing list, there seems to be some support to make OP_RETURN more attractive to discourage the "misuse" of the Taproot data storage features. So it is possible such a BIP could get support.
centralized cryptocurrency might have such feature.
The "make OP_RETURN completely ignorable" feature would not need any centralized elements. Not even a hypothetical "notify and take down" system would need that, although of course those who do the "notifications" need to be trusted from the full node operators.
But full node operator is unlikely to use such feature
Why? I guess node operators located in countries with strict legislation on illegal content, fearing legal action, would like to use that.
and feature specific for Bitcoin Core doesn't need BIP.
It would be a protocol change, because currently OP_RETURN messages cannot (afaik) be completely ignored, they can only be pruned once validation of the block where they're located has been done. So it would need a BIP. Bitcoin Core as the reference implementation would be the first to have to support it. (Don't really get what you mean here.)
That's possible, but don't forget it's more expensive since they doesn't get witness "discount".
Somebody who can expect enormous profits shorting the price down can afford that, even if he had to pay 100 times the Taproot fee.
By the way, there seem to be currently
significantly less inscriptions than last week. However, the average size has risen a bit, so the impact on block size (while smaller than before) is still quite high.