- Identifiers are now evaluated as variables instead of being plain values
- Constants like `left` or `bold` are stored as dynamic values containing the respective rust types
- We now distinguish between arrays and dictionaries to make things more intuitive (at the cost of a bit more complex parsing)
- Spans were removed from collections (arrays, dictionaries), function arguments still have spans for the top-level values to enable good diagnostics
- Better tokenization test coverage.
- Suffix testing: Each test case is tested with many different suffixes to ensure correct token ends.
- Improves expression parsing (fixes#3).
Laurenz