mtmd : Fix 32-bit narrowing issue in export-lora and mtmd clip

[kiwi142857]

Summary

Fixes narrowing conversion errors when building on 32-bit platforms due to implicit conversion from long long to size_t in initializer lists.

Under C++11 rules, non-constant expressions in initializer lists that narrow types will cause compile errors. This affects both export-lora.cpp and clip.cpp:

gguf_get_n_tensors(...) * ggml_tensor_overhead(); // returns long long

On 32-bit platforms, this cannot be implicitly converted to size_t (unsigned int) in an initializer list.

Fix

Added static_cast<size_t> to silence narrowing warnings and ensure type safety across 32/64-bit platforms.

Files affected

•tools/export-lora/export-lora.cpp
•tools/mtmd/clip.cpp

Notes

•This change is non-functional: it only impacts build correctness.
•Confirmed to compile cleanly on both 32-bit and 64-bit targets using clang.