Add fuzzer for json_decode module

[AdamKorcz]

Fuzzes the CPython _json C module (Modules/_json.c) through JSONDecoder.decode() and JSONDecoder.raw_decode(), dispatched per input via FuzzedDataProvider. Input bytes are decoded as latin-1 so every byte value maps to a distinct code point, preserving the full 0–255 byte space at the parser boundary — in contrast to json.py, which feeds UTF-8 with errors="replace" and collapses any invalid sequence to U+FFFD, sharply shrinking the effective input space. It also reaches raw_decode()'s trailing-data position reporting that json.py never calls, and drops the dumps/loads roundtrip to focus purely on decoder hardening rather than re-encoding already-valid objects.