it is an output containing a smart contract that creates a "lock" that can be "unlocked" when spending that output
How is that different from any token? If you own "1 ALT", then it also has a smart contract. It also can be unlocked. By using zero amount, you can just disable amount validation on Bitcoin, but all other properties are preserved. To move it, you still need to provide a matching input, exactly in the same way as for any tokens on other chains.
Since that won't be part of the Bitcoin protocol, it can not be considered a "token".
If there are N parties, then N-of-N multisig can be always used as a base (and nicely compressed by using Taproot address with Schnorr signatures). And then, you can create as long Script as you want, it can be expanded into many transactions, and since Taproot, the upper limit is 4 MB (or something like 10% of that, if you want to stick with standard transactions). The Script is powerful enough to express any logical operation: by using OP_TRUE, OP_FALSE, OP_EQUAL, OP_NOT, OP_IF and OP_ENDIF, you can build absolutely everything. If it will be too large, then you can focus more on optimization, use opcodes like OP_ADD to count things on 32-bit values, instead of doing it bit-by-bit, or split it between many transactions. So yes, you can for example create some output, that could be unlocked by Ring signatures from Monero. It will be another attack vector on Bitcoin, it will be too complex, it will take a lot of space, but the Pandora box is opened, and I think it is a matter of time.
Because it is for testing and there is no guarantee that we don't decide to move to testnet4 some day, discarding the whole testnet3 chain.
I doubt we will ever see testnet4. Why?
1. There were a lot of halvings on testnet3. If you want to test the same things on testnet4, then you need a lot of history, but if your starting point is the Genesis Block, then it would take a lot of time and power to produce such chain.
2. Core Developers moved to signet, instead of creating testnet4. They have no reason to create another testnet, when they have a more stable chain, without any blockstorms, that can be fully controlled, and easily turned into testnet, just by changing "signetchallenge" into "OP_TRUE".
3. Testnet3 existed for many years, it survived longer than many altcoins, I doubt node operators will abandon that chain, even if it will be removed or resetted in some new client version.
4. If resetting the chain is a good thing, and we need a test network, where old transactions and blocks are discarded after some time, then it should not be triggered manually, but automatically. Also, there is no reason to improve testnet3, if the same improvements could be applied to signet instead.
As you already know, testnet coins are worthless so there won't be any incentive to keep them, which means migration to version 4 won't be met with that much resistance.
You are talking about the chain, that is stronger and older than many altcoins.