Initial exploration around JIT

[ClementWalter]

An initial standalone test for using JIT compilation of the EVM bytecode.

In this test, I've implemented few opcodes that I felt characteristic of the needs of the real EVM bytecodes, hopefully not missing any of the real difficulties with implementation of

• PC opcode where it required to read from the return return fp - 3 to get the pointer to the beginning of the segment
• JUMP like opcode where one skips the next n opcode
• PUSH like opcode where one read value from the next opcodes and jump to the next opcode
• RET like opcode where the execution halts
• CALL like opcode where the opcode does jit internally to trigger a subcall and returns its value

Overall, I feel that with very few changes to the opcodes as they are, and a lot of simplification for the syscall and at the interpreter level, we can migrate to jit, which will also save steps (to be estimated) but at least all the input/output argument copying + control flow management

[ClementWalter]

blocked because this needs a fix in the cairo vm

[Eikix]

Should we close this PR for now and come back later to it?
I can also convert it into a P2 issue