v1.2.4

Fixes a crash on non-ASCII input and accepts such identifiers like SQLite.

Typing an identifier with a non-ASCII character (accented, Cyrillic, …) — e.g. café or привет — previously crashed consumers such as SQLite ORM Studio: the tokenizer's "unexpected character" error embedded a lone UTF-8 byte, and serializing that invalid-UTF-8 message threw and aborted the app (#25).

The tokenizer now treats any byte >= 0x80 as an identifier character, matching SQLite (which never decodes UTF-8). Non-ASCII identifiers tokenize and flow through codegen; the C++ struct/column names are sanitised (non-ASCII bytes → _) while the original name is preserved in the generated make_table/make_column strings (#26).