Improve P2E transfer

[yrong]

Context

For transfers from a non-system parachain to Ethereum, since we perform a full dry-run check at each node/hop, the previous fallback XCM — which sent assets back to the source chain from AH when execution failed on AH — is no longer necessary.

This PR removes that fallback, slightly reduces fees, and has been tested and verified with a V2 transfer from Bifrost, which worked as expected.

https://app.snowbridge.network/activity#0x7a788c60bd72580111a5b4cbab1bed244f8f6e134d78a97310afadb7eae39d01

[yrong]

Tested that the V2 transfer does not work on Acala, as the runtime does not permit PolkadotXcm.execute, as shown here:
https://github.com/AcalaNetwork/Acala/blob/eb004a3cdb18cae68875e080e1b731c996dd3dc1/runtime/acala/src/lib.rs#L238

[claravanstaden]

Does it hurt to keep "return to sender" in? I think its a nice fallback, but @alistair-singh would have the final say on this :)

[yrong]

Does it hurt to keep "return to sender" in? I think its a nice fallback

It will help users save some costs. By the way, I left the V1 path untouched. For V2, since we already have an optional claimerLocation — typically the original sender —it is internally handled as SetHints(Claimer), which follows more of an XCM V5–style approach.

[alistair-singh]

@claravanstaden Since we are only doing this for v2 paths only where dryRun is guaranteed i think its fine to do this. The return to sender code costs a lot by the time we apply padding, and sometimes i think we apply padding on padding so its way too expensive. Its also error prone and not 100% guaranteed to return to sender, it can sometimes fail, such as not having enough fees. So I am for this both simplifying code and lowering cost.