Use XCM benchmarks weights results instead of fixed estimations

[NachoPal]

Currently, XCM weights are fixed and taken from here: https://github.com/paritytech/cumulus/blob/master/polkadot-parachains/statemine/src/xcm_config.rs#L126

Weights from the benchmarks results are only used in Westend:
https://github.com/paritytech/polkadot/blob/master/runtime/westend/src/xcm_config.rs#L111
https://github.com/paritytech/polkadot/blob/master/runtime/westend/src/xcm_config.rs#L137

I am wondering if this is intentional and if benchmarks weights can be applied also at least for Westmint, Statemine and Statemint.

cc @shawntabrizi @KiChjang

[shawntabrizi]

XCM benchmarks are basically in beta. I would be happy for you to check if they run here and integrate them if they do.

If they don't, would be good to see why.

[NachoPal]

XCM benchmarks are basically in beta. I would be happy for you to check if they run here and integrate them if they do.

If they don't, would be good to see why.

Why exactly the XCM benchmarks are still in beta. What are your concerns about them and how do you think they might negatively impact Statemine/mint runtimes?

Is it just about inaccurate weight values?

[shawntabrizi]

No, its just that XCM is very generic and configurable, and that the benchmarks is not proven to work 100% of the time. You may run into errors, and we just need to catch those and fix them.

The only reason they have not been integrated is that it takes someone to do it, and I have been busy with other stuff.

@KiChjang has been tasked with this, but he is also very busy. So, if you feel you can jump in, would be very happy to have you try to integrate them.

[KiChjang]

I'm actually trying to fix the remaining XCM benchmarks here, but the compile-test-code cycle takes a really long time when it comes to benchmarks, but once it's done, I'll be taking over Shawn's XCM benchmark PR for Kusama and getting it to the finish line. By then I may be able to look at how we do the same benchmark with the asset parachains, if nobody else has gotten to it before me.

[gilescope]

We should be able to do this for statemine/t now that paritytech/polkadot#4442 has landed for the relay chain.