llama.cpp Release b11063 Fixes Invalid UTF-8 Handling in PEG Parser

llama.cpp release b11063 introduces a patch to handle invalid UTF-8 sequences gracefully within the Abstract Syntax Tree (AST) of its Parsing Expression Grammar (PEG) parser module. The update ensures the parser returns maximal subparts for malformed bytes according to Unicode standard recommendations. This fix enhances parser robustness when enforcing structured output grammars (such as JSON) during LLM generation. It prevents system crashes or unexpected parsing failures when handling raw or corrupted text inputs containing invalid character encodings. Implemented in PR #29161 under the common/peg module, the patch removes strict decoding arguments and standardizes bad byte sequence replacements. Pre-built binary artifacts for macOS, Linux, Windows, Android, and iOS were released alongside the code update.