Add OpenBLAS for NEON-optimized matrix operations

Added OpenBLAS to vcpkg dependencies to accelerate Armadillo's
Cholesky decomposition with hand-tuned ARM NEON kernels.

Expected 2-5x speedup for clutter filter Cholesky operations
(450ms -> 90-180ms target).

Updated Dockerfile to copy OpenBLAS runtime libraries.

Author: Sol455

Dockerfile

@@ -138,6 +138,10 @@ COPY --from=blah2_env /usr/local/lib/libfftw3.so* /usr/local/lib/
 COPY --from=blah2_env /usr/local/lib/libfftw3f.so* /usr/local/lib/
 COPY --from=blah2_env /usr/local/lib/libfftw3_threads.so* /usr/local/lib/
 COPY --from=blah2_env /usr/local/lib/libfftw3f_threads.so* /usr/local/lib/
+
+# Copy OpenBLAS libraries from vcpkg build stage
+COPY --from=blah2_env /opt/blah2/lib/vcpkg_installed/*/lib/libopenblas* /usr/local/lib/
+
 RUN ldconfig
 
 # Copy the compiled binary from build stage

lib/vcpkg.json

@@ -1,11 +1,12 @@
 {
   "name": "blah2",
   "version": "1.0.0",
-  "dependencies": [ 
+  "dependencies": [
       { "name": "catch2", "version>=": "3.4.0" },
       { "name": "rapidjson", "version>=": "1.1.0" },
       { "name": "asio", "version>=": "1.28.0" },
       { "name": "cpp-httplib", "version>=": "0.12.2" },
+      { "name": "openblas", "version>=": "0.3.0" },
       { "name": "armadillo", "version>=": "12.0.1" },
       { "name": "ryml", "version>=": "0.5.0" }
   ],