Gather for an EIP: Useful Bit manipulation instructions

[axic]

@chfast

CLZ seems to have the complexity of SHL, i.e. it is messy. There is a lot of handling of individual words being zero. I have not seen constant-time implementation yet. Having lzcnt instruction available should help a bit.

@Vectorized

For the implementation of CLZ, I’d propose a gas cost of 3
(to keep in line with the gas cost of shl, shr, not, xor, etc).

Just keep in mind, that while from an EVM code perspective they seem simple, due to the EVM operates on 256-bit big endian words and most CPUs operate on 64-bit little endian words, this becomes slightly more complex to implement, and the cost should reflect that. One could argue that not/xor is overpriced compared to shl/shr.