glob.nx

glob.nx -- POSIX-style glob pattern matching. Supports: ? match exactly one byte (any byte except null) * match zero or more bytes (greedy, with backtrack) [abc] match any byte in the set [a-z] match any byte in the inclusive range [!abc] negated set (not any of abc) \\x literal x (escapes *, ?, [, \) anything literal byte match Used by: fs.nx dir-walk (list files matching *.nx), nxmake.nx dependency inputs, ignore files (.gitignore-style patterns). This is NOT a full regex -- just POSIX fnmatch's basic set. Don't try to encode complex patterns; use a real regex engine when they appear (roadmap: runtime/re.nx, TBD). Invariants: G1 Pure function: no memory allocation, no syscalls; runs purely on the input strings. Safe in tight loops. G2 Linear time in (pattern_len * text_len) worst case -- no catastrophic backtracking because `*` backtrack is bounded by text length. G3 Null-byte terminator aware: both pattern and text stop at '\0'. Callers with embedded-null bytes should pre- sanitize. G4 Character classes support byte ranges only; no Unicode codepoint classes. UTF-8 multibyte treated as opaque bytes (still matches correctly for byte-exact input).