Reject `BLOCKHASH`

[chfast]

The EIP-2935: Serve historical block hashes from state possibly modified the BLOCKHASH instruction and provides enhanced alternative in form of a system contract.

Considering there is going to be better way for accessing block hashes the BLOCKHASH instruction should be rejected in EOF.

Pros

• single way of accessing block hashes
• BLOCKHASH instruction is weird
• the less instructions the better

Cons

• BLOCKHASH instruction provides uniform way of accessing this information across EVM chains (not every chain may have the system contract, or have it before EOF)
• BLOCKHASH is cheaper (at least before Verkles)
• BLOCKHASH is smaller code

[jochem-brouwer]

I do not understand the motivation here. The only way I see how BLOCKHASH has changed is that there now is a bigger window to lookup hashes. If we remove it then you have to do this system call which is definitely more expensive than BLOCKHASH.

[chfast]

The goal is to remove obsolete/deprecated features of EVM in EOF so the EOF is simpler.

[chfast]

If the decision is "no change" we should still write some EIP "things considered" section.