With this proposal/improvement, i wonder about these things :
1. Would this affect physical merchants who accept 0-confirmation, since transaction propagation will be longer and merchant/user might have wait a bit longer while there's queue?
No, in 2 ways.
1st way: Transaction propagation using BIP 156 (i.e. dandelion tx relay) will still be very fast in relative terms, so it won't make any difference in the real world
2nd way: Accepting 0-confirmation transactions will be just as risky (and inadvisable) as they are without dandelion
2. Would this affect block size/weight limit size increase in future?
No
3. Is using Tor/I2P/Kovri better/simpler solution?
It's gonna be easier to use BIP 156, it doesn't require the user to do anything (Tor requires some extra setup). But it won't work 100% of the time for a while yet, as BIP 156 (dandelion) needs compatible nodes on the Bitcoin network for it to work; it's currently planned for inclusion in version 0.18.0. Only 0.18.0+ nodes will be able to propagate transactions the dandelion way (unless it gets backported to older versions, which seems unlikely as it's a new feature), so it will take some time before you can expect every transaction to be relayed using this technique. I imagine the protocol will be implemented so that BIP 156 nodes prefer to relay only to other BIP 156 compatible nodes, but the implementation isn't finished AFAIK.