feat[venom]: remove dloadbytes and sha3_64 venom instructions

[harkal]

What I did

Remove dloadbytes and sha3_64 venom instructions, and updated the translator to emit directly lower level instructions.

How I did it

How to verify it

Commit message

This commit removes the higher level `dloadbytes` and `sha3_64` venom instructions.

Description for the changelog

Cute Animal Picture

[charles-cooper]

I think that we want dload and dloadbytes actually, they might be compiled differently on other architectures (which handle immutables differently)

[harkal]

I think that we want dload and dloadbytes actually, they might be compiled differently on other architectures (which handle immutables differently)

We can always add them when we need them -and we know the exact requirements-

[charles-cooper]

i think we need them, zksync handles immutables as storage variables. dload loads from a different location than code

[harkal]

Since zkSync currently does not support Venom, merging this will not cause any disruptions. Furthermore, dstore was never part of Venom to begin with. To properly support zkSync or other architectures, we would need a command-line parameter for architecture selection. A less invasive approach would be to treat it as a special version of the EVM, although this isn't ideal. Venom will utilize architecture-specific information to decide which opcodes are permissible (or required) and even to choose more efficient opcodes on architectures where gas consumption varies, etc.

[harkal]

Added an issue about it #4326

[harkal]

Check #4357

[charles-cooper]

i think we could even keep this open, since the other PR is more concerned with the frontend generating the instructions in the first place

[charles-cooper]

the dloadbytes portion is superseded by #4410. let's leave this open for the sha3_64 change, although i think it can help with cse and instruction reordering