The current inline alignment is very broken and leads to lots of subtle weirdness. Getting rid of it simplifies the stack's interface a lot. At a later point either:
- inline alignment will be added back in a better way, or
- all nodes will be able to expand or align themselves, meaning that the stack's children take care of their alignment
- Templates scope state changes
- State-modifying function operate in place instead of returning a template
- Internal template representation contains actual owned nodes instead of a pointer to a syntax tree + an expression map
- No more wide calls
- Makes errors fatal, so that a phase is only reached when all previous phases were error-free
- Parsing still recovers and can produce multiple errors
- Evaluation fails fast and can thus produce only a single error (except for parse errors due to an import)
- The single error that could occur during execution is removed for now
- Removes Value::Error variant
