fix: resolve unsigned integer initialization warnings in gguf.cpp

[skrandy]

Fix warnings caused by initializing unsigned integers (uint32_t n_dims and uint64_t size) with negative values.

• Changed uint32_t n_dims = -1 to 0 in tensor shape parsing.
• Changed uint64_t size = -1 to 0 in string reading logic.

These values are overwritten by subsequent gr.read(...) calls, so using 0 as an initial placeholder eliminates unsigned integer underflow warnings without affecting runtime behavior.