Unify fee modifiers

[notlesh]

What does it do?

This accomplishes two things:

1. Remove the floating congestion-based fee modifier previously used for pallet transaction-payment
2. Unify fee multipliers for both Ethereum and Substrate transactions

Note that transaction-payment doesn't actually use the MultiplierUpdate trait outside of its integrity_test hook, it actually uses the Convert trait at runtime:

https://github.com/paritytech/substrate/blob/9461b2de04210c6c193726a745c3ec6552b4ce9f/frame/transaction-payment/src/lib.rs#L344

The previously used TargetedFeeAdjustment impl used these values, but without that they are completely unused at runtime (please review).

I believe, then, that this will work with fixed fees for now and should also be fine in the future when EIP-1559 congestion-based fee modifier is properly implemented.

This is closely related to #1353 but should be fine if done independently.

⚠️ Breaking Changes ⚠️

⚠️ All substrate-based fees will be impacted by this
⚠️ Some txns in the mempool prior to this change (runtime upgrade) may become invalid
⚠️ RPC calls to estimate fees will return different values
🧷 This only affects moonbase runtimes currently

TODO:

• Bounds checking (U256 -> U128 -> FixedU128)
• DRY (move to common)
• Tests, esp. that the values satisfy integrity_test

[notlesh]

This test sucks. Basically, I tried to exploit the fact that remark's weight fn returns its length, but after some debugging this appears to be the encoded length, not the actual remark length.

#[pallet::weight(T::SystemWeightInfo::remark(_remark.len() as u32))]

The overhead increases as the remark grows in size, probably because it takes more bytes to describe the size of a larger vec.

I think the test does its job, but it's pretty ugly. I'd like to find a cleaner way to accomplish the same thing.

[notlesh]

Closing this in favor of #1765