UPSTREAM PR #17949: cmake: link ws2_32 for MinGW/w64devkit builds in cpp-httplib

[loci-dev]

Mirrored from ggml-org/llama.cpp#17949

Small linker fix to allow the procedure described in build.md to work correctly using w64devkit.

Missing link found with Claude and corrected with Gemini.

[loci-review]

Explore the complete analysis inside the Version Insights

Performance Analysis Summary: PR #530

Analysis Scope: Build system modification for MinGW/w64devkit Windows builds
Change Type: CMake linker configuration
Files Modified: 1 (vendor/cpp-httplib/CMakeLists.txt)

This PR adds ws2_32 library linkage for MinGW Windows builds of cpp-httplib. The change is a build system fix with no source code modifications. Power consumption analysis shows 0.0% change across all binaries including libllama.so, llama-run, and libggml components. No function-level performance data indicates measurable differences between versions. The modification affects only HTTP server compilation and does not touch inference paths, tokenization logic, or GGML tensor operations. Functions responsible for inference performance (llama_decode, llama_encode, llama_tokenize) remain unchanged. No impact on tokens per second throughput.