Remove `delete_existing` option; Introduce archive

[azrv]

Remove delete_existing run option, instead archive memory and workspace on every run;

archive is a step since some flows don't generate code and therefore should not archive whatever

Relates to https://github.com/AntonOsika/gpt-engineer/issues/388

[AntonOsika]

Great work!

Added some comments.

Mainly I think we shouldn't let archive be a "step" but we just call it.

Also, feel free to delete the "run_id" functionality.

[azrv]

Great work!

Added some comments.

Mainly I think we shouldn't let archive be a "step" but we just call it.

Also, feel free to delete the "run_id" functionality.

@AntonOsika

1. My main concern about archiving NOT being a step is there are "flows" that don't generate code (e.g. EXECUTE_ONLY, EVALUATE) and therefore should not archive things by default. How do we make sure to skip archiving in those cases? While working on PR I realised archiving functionality complicates code a lot and I'm no sure it provides us any good. What's inherently wrong with delete_existing?

2. Sure! Could you please elaborate on what run_id is I can't seem to find it (not so familiar with the codebase yet) git grep run_id gives nothing

[AntonOsika]

Merging!

[AntonOsika]

Thanks for the submission @azrv !

Do you want to join discord and DM me there? Happy to chat about future improvements!

[AntonOsika]

Already cleaned up run_id (was called run_prefix)