ossfuzz and Yul interpreter use ast over block

[clonker]

Changing the interface of the Yul interpreter to use an instance of AST over individual block and dialect. With numeric IDs, the AST will also contain the labels, so this makes the overall interface leaner.
On that note I realized that yul interpreter tests are parsed potentially with eof (in CI, for ex.) but are interpreted without.

[cameel]

I realized that yul interpreter tests are parsed potentially with eof (in CI, for ex.) but are interpreted without.

I don't think it's really a problem. EOF largely leaves semantics the same, so most of the tests that pass should still make sense on it. The ones using opcodes that were added or removed won't work, but they are already disabled with bytecodeFormat: legacy.

I'd rather keep it like that and only disable tests on a case-by-case basis. The default state should be that things get tested both with and without EOF and you have to manually exclude them for things that are not yet implemented. There should be natural pressure for the amount of such tests to shrink over time.

[clonker]

If you put it like that it makes complete sense to do it like that. I was coming from the point of view that by essentially 'enabling' EOF in the interpreter (i.e. no longer hardcoding the dialect to be EOF-free) changes the current behavior and might make it go unnoticed that there's still places that need to be EOFed. Although all it does is increase testing surface which isn't a bad thing :)
In yulrun it is still hardcoded to not use EOF but I have added the assert anyways, for good measure. That way it can't slip through the cracks.