PaysFee for DispatchInfo

[xlc]

This allow more flexible fee logic such as adjust fee based on DispatchClass.

I want to have operational transactions have weight but no fees.

[kianenigma]

I want to have operational transactions have weight but no fees.

Fair and indeed a problem right now that we should fix.

I will think a bit about this and maybe and see if I can come up with a more general way of expressing fee logic in #[weight] without making it too complicated. Ideally, I want to get rid of a horrible function pay_len_fee() which is total crap as well. But this PR also seems like a good start. Thanks :)

[kianenigma]

Needs further test fixing as well.

[kianenigma]

@xlc how about this:

• get rid of pay_length_fee().
• A third trait that all weight representatives must implement named Trait PaysFee { fn pays_fee() -> bool { true; } } + a blanket impl for all structs that have the other two traits (WeighData, ClassifyDispatch).
• new filed in DispatchInfo named pays_fee: bool which is being set by dispatch.rs.

Then, whoever calls get_dispatch_info() know this and it can be expressed in the #[weight] tag. SimpleDispatchInfo does not show it atm but we can readily make it possible.

[kianenigma]

just a minor not here and there and then good to go. As mentioned, it won't be the end of the deal and I want to retire SimpleDispatchInfo soon, which will happen in follow ups.

The only logical difference of this PR is that a weighted operational transaction (we don't have that many of these actually, most FreeOperational) now pay the length fee as well.