However, the idea with BitVM is that any cryptosystem could be emulated.
Yes. But it is costly, if you have to put all of that on-chain. It is much better to just have some external proof, and only commit to it on-chain. For example: something like this sudoku puzzle:
https://github.com/zcash-hackworks/pay-to-sudokuImagine how more complex everything would be, if you would push the whole NxN sudoku puzzle inside the input script.
The Bitcoin developers would have some time more to decide which algorithm is the best one
No matter what you pick, it should be deployed on some kind of test network first anyway. So, no matter if it would be a soft-fork, or a long Script, the way of deployment is similar.
I'm however not sure how much must be posted on-chain in the case a spend is contested.
If you have some external proof, then in most cases, you cannot spend the coin on-chain, if you don't know the solution.
And if some different elliptic curve is used, then you can use a DLEQ proof:
https://groups.google.com/g/bitcoindev/c/MezoKV5md7sif BitVM can/should be used for this purpose
I guess not, because if you have to put a lot of data on-chain, then it is an equivalent to shrinking the maximum block size. And those limits will not be lifted in the nearest future, because it would only open a wide door for some spam, and the amount of real coin usage will be left unchanged (or even some users will switch into other networks, as a result).