Drop OSEK support

[michael-schwarz]

For reference, Goblint v1.1.1 still contains all the OSEK analyses, but we are removing them to ease the maintenance burden. Should you be interested in these OSEK analyses, feel free to reach out!

TODOs:

• Remove interrupt edges
• See what sort of hacks we can get rid of now

Closes #735

[sim642]

There's also something in scripts/old/osek.

And I checked that postsub can finally be removed now! Disappointingly, presub is still used by ARINC.

[michael-schwarz]

And I checked that postsub can finally be removed now!

Should we do this here or save it for another PR?

[sim642]

Should we do this here or save it for another PR?

Maybe separate, because I think we could actually get rid of presub's only usages in ARINC as well. There it's just being used to get the context of base analysis (only!) and compute its hash. One possibility might be to even compute the hash of ctx.control_context instead, which would also account for other analyses. It's an easy change to make, but should probably be tested, since it could also completely break ARINC support (if it isn't already).
Or the safer alternative would be a query instead, which works because queries are answered on the pre-state anyway.

If we manage that, then the whole idea of topological sorting of analyses and not having dependency cycles becomes unnecessary (they had to be sorted for postsub to be able to access the post-state of dependent analyses). AFAIK we somewhere currently omit declaring a dependency, because it would cause a cycle, but we actually have a dependency that another analysis answers some query. With joint-querying it's not really necessary to order the analyses.