Increase compile performance by removing Intermediate/Implementation #65
Description
Currently, every time you call #step the entire Sequence is recompiled into Intermediate and Implementation, and then passed to Activity::Schema::Compiler.
If Activity::Schema had a simplified build interface that receives something that's almost a Circuit plus the required data such as start_task, we could throw away a lot of code in dsl, speed up compilation by probably factor 2-3 and move Intermediate code to trailblazer-workflow, as that's why this intermediate layer got invented for.
Once this is done, move old "snippet/activity/low_level" docs to wherever this layer ends up.