Draft updates 2

[Vectorized]

• ERC20 and native token spend permissions.

Todo:

• Write tests.
• EntryPoint create2 address.
• Ability to let delegation approve more than one EntryPoint (tied to current delegation codehash). This allows for the delegation to possibly switch to a more performant, but abi equivalent EntryPoint in the future.

Thoughts:

• The EntryPoint is quite a complex beast, due to it's need for gas introspection and performance, just like the 4337 EntryPoint. I think in the end state, a good EntryPoint will be written in very optimized inline-assembly. Also, the main bottleneck in Base's gas costs by far is L2 gas.

[gakonst]

WDYT about the AA folks exposing the validation logic in a separate function? Is it worth it?

I think there is a benefit because it'd let us build a Reth executor which prefetches session keys / verifies sigs in parallel, which could improve performance

I know it introduces the pre verification gas stuff which IMO should not be exposed to devs so I'd like to avoid that if we thought doing the validation / execution separation was worth it.