If I recall, implementing zk-snarks on smart-contracts would massively increase the resource burden on the Ethereum blockchain. This is why they won't be implementing these until they have achieved parallelization, ergo-added sharding to help manage the additional load. However, by adding both sk-snarks and sharding, there will likely be no performance increase which is the entire purpose of ETH 2.0.
It makes sense to scale the ETH blockchain first, before adding Zero-Knowledge capabilities to smart contracts. If successful, then the ETH blockchain will be a game-changer as both ERC-20 and ERC-721 tokens will gain a full-fledged privacy solution for mainstream businesses and corporations to use. The new ZK-STARKs technology aims to take less space (as it's compressed) than the current ZK-SNARKs used by Zcash and some other privacy-oriented cryptocurrencies. But it might take a couple of years from now before Ethereum adopts Zero-Knowledge Proofs for its own benefit. Hopefully, the next iteration of Ethereum (v2.0) would bring us Sharding and PoS to pave the way for privacy-oriented smart contracts. I wouldn't be surprised if another Ethereum fork takes birth in order to address the issue of privacy (in case ETH doesn't take the anonymity route).
In the meantime, we should focus more on ETH becoming widely used in the mainstream world than anything else. Then, we could talk about privacy and other optional features. Just my opinion