ok then why aren't people doing that then? go ahead and answer. is it because it is too expensive or inconvenient or they just don't know about it.
The truth is that this has been done several times since 2013 or 2014, but never a "fad" developed around these early protocols. In fact, it was an intelligent move of the developers to make OP_RETURN standard in 2014, because as I wrote the earlier protocols clutter the UTXO set, while OP_RETURN outputs are not stored there (they can be safely pruned by nodes - @franky1 incoming saying "these are not full nodes" 3... 2... 1... 
).Stampchain tried to introduce a similar protocol called SRC-20 recently, embedding the tokens into "normal-looking" multisig transactions. Fortunately (for the UTXO set) they failed to get any adoption, probably due to Runes being already discussed at that time they launched.
see if i was the one that designed bitcoin i would have made it impossible for all these type of things to happen. my bitcoin would be a bit lean and mean. and if people wanted to pollute the utxo set then they would pay a hefty price for doing so. because there would be no data storage component.
OK, so how is your plan?The challenge is that you can actually not allow anything which can be used for arbitrary data. For example, probably publicly visible fields for values with a large number of digits could be enough. Public key hashes too, like I demonstrated in the post I linked above.
You could have forked Grin or Monero, but these protocols were developed years (Monero: 2014, Grin: 2018) after Bitcoin was. So you would have had to travel in time.
And even these do allow some inscription mechanisms but you would have to link several transactions so they would be effectively more expensive than "payment" transactions. For example, for Monero Mordinals exists, and for grin at least a proof of concept was shown on Github.