[eigen3] Use upstream branch 3.4, instead of master

[luncliff]

Changes

• Not using master branch. Prepare the upcoming 3.4.1 release.

References

• [eigen3] Update to 2024-08-01 #224
• https://gitlab.com/libeigen/eigen/-/tree/3.4?ref_type=heads

Triplet Support

All triplets in the registry

[luncliff]

Failure in tensorflow-lite

[232/256] D:\Android\ndk\27.1.12297006\toolchains\llvm\prebuilt\windows-x86_64\bin\clang++.exe --target=aarch64-none-linux-android26 --sysroot=D:/Android/ndk/27.1.12297006/toolchains/llvm/prebuilt/windows-x86_64/sysroot -DCPUINFO_SUPPORTED_PLATFORM=1 -Dtensorflow_lite_EXPORTS -DTENSORFLOW_CORE_PLATFORM_FLOAT8_H_ -DTENSORFLOW_TSL_PLATFORM_FLOAT8_H_ -IC:/Users/luncl/Documents/vcpkg-registry/buildtrees/tensorflow-lite/src/v2.17.0-6beb5bd9c1/third_party/xla/third_party/tsl -IC:/Users/luncl/Documents/vcpkg-registry/buildtrees/tensorflow-lite/src/v2.17.0-6beb5bd9c1/third_party/xla -IC:/Users/luncl/Documents/vcpkg-registry/buildtrees/tensorflow-lite/src/v2.17.0-6beb5bd9c1 -IC:/Users/luncl/Documents/vcpkg-registry/buildtrees/tensorflow-lite/src/v2.17.0-6beb5bd9c1/tensorflow/lite/c -isystem C:/Users/luncl/Documents/vcpkg-registry/installed/arm64-android/include/eigen3 -isystem C:/Users/luncl/Documents/vcpkg-registry/installed/arm64-android/include -isystem C:/Users/luncl/Documents/vcpkg-registry/installed/arm64-android/include/gemmlowp -g -DANDROID -fdata-sections -ffunction-sections -funwind-tables -fstack-protector-strong -no-canonical-prefixes -D_FORTIFY_SOURCE=2 -Wformat -Werror=format-security -frtti -fexceptions  -fPIC   -fno-limit-debug-info    -std=gnu++20 -fPIC -DEIGEN_NEON_GEBP_NR=4 -DNNAPI_VERBOSE_VALIDATION -DTFLITE_BUILD_WITH_XNNPACK_DELEGATE -DXNNPACK_DELEGATE_ENABLE_QS8 -DXNNPACK_DELEGATE_ENABLE_QU8 -DXNNPACK_DELEGATE_USE_LATEST_OPS -DXNNPACK_DELEGATE_ENABLE_SUBGRAPH_RESHAPING -DTFLITE_WITH_RUY -Wno-deprecated-declarations -pthread -MD -MT CMakeFiles/tensorflow-lite.dir/nnapi/nnapi_handler.cc.o -MF CMakeFiles\tensorflow-lite.dir\nnapi\nnapi_handler.cc.o.d -o CMakeFiles/tensorflow-lite.dir/nnapi/nnapi_handler.cc.o -c C:/Users/luncl/Documents/vcpkg-registry/buildtrees/tensorflow-lite/src/v2.17.0-6beb5bd9c1/tensorflow/lite/nnapi/nnapi_handler.cc
FAILED: CMakeFiles/tensorflow-lite.dir/nnapi/nnapi_handler.cc.o 
D:\Android\ndk\27.1.12297006\toolchains\llvm\prebuilt\windows-x86_64\bin\clang++.exe --target=aarch64-none-linux-android26 --sysroot=D:/Android/ndk/27.1.12297006/toolchains/llvm/prebuilt/windows-x86_64/sysroot -DCPUINFO_SUPPORTED_PLATFORM=1 -Dtensorflow_lite_EXPORTS -DTENSORFLOW_CORE_PLATFORM_FLOAT8_H_ -DTENSORFLOW_TSL_PLATFORM_FLOAT8_H_ -IC:/Users/luncl/Documents/vcpkg-registry/buildtrees/tensorflow-lite/src/v2.17.0-6beb5bd9c1/third_party/xla/third_party/tsl -IC:/Users/luncl/Documents/vcpkg-registry/buildtrees/tensorflow-lite/src/v2.17.0-6beb5bd9c1/third_party/xla -IC:/Users/luncl/Documents/vcpkg-registry/buildtrees/tensorflow-lite/src/v2.17.0-6beb5bd9c1 -IC:/Users/luncl/Documents/vcpkg-registry/buildtrees/tensorflow-lite/src/v2.17.0-6beb5bd9c1/tensorflow/lite/c -isystem C:/Users/luncl/Documents/vcpkg-registry/installed/arm64-android/include/eigen3 -isystem C:/Users/luncl/Documents/vcpkg-registry/installed/arm64-android/include -isystem C:/Users/luncl/Documents/vcpkg-registry/installed/arm64-android/include/gemmlowp -g -DANDROID -fdata-sections -ffunction-sections -funwind-tables -fstack-protector-strong -no-canonical-prefixes -D_FORTIFY_SOURCE=2 -Wformat -Werror=format-security -frtti -fexceptions  -fPIC   -fno-limit-debug-info    -std=gnu++20 -fPIC -DEIGEN_NEON_GEBP_NR=4 -DNNAPI_VERBOSE_VALIDATION -DTFLITE_BUILD_WITH_XNNPACK_DELEGATE -DXNNPACK_DELEGATE_ENABLE_QS8 -DXNNPACK_DELEGATE_ENABLE_QU8 -DXNNPACK_DELEGATE_USE_LATEST_OPS -DXNNPACK_DELEGATE_ENABLE_SUBGRAPH_RESHAPING -DTFLITE_WITH_RUY -Wno-deprecated-declarations -pthread -MD -MT CMakeFiles/tensorflow-lite.dir/nnapi/nnapi_handler.cc.o -MF CMakeFiles\tensorflow-lite.dir\nnapi\nnapi_handler.cc.o.d -o CMakeFiles/tensorflow-lite.dir/nnapi/nnapi_handler.cc.o -c C:/Users/luncl/Documents/vcpkg-registry/buildtrees/tensorflow-lite/src/v2.17.0-6beb5bd9c1/tensorflow/lite/nnapi/nnapi_handler.cc
In file included from C:/Users/luncl/Documents/vcpkg-registry/buildtrees/tensorflow-lite/src/v2.17.0-6beb5bd9c1/tensorflow/lite/nnapi/nnapi_handler.cc:15:
In file included from C:/Users/luncl/Documents/vcpkg-registry/buildtrees/tensorflow-lite/src/v2.17.0-6beb5bd9c1/tensorflow/lite/nnapi/nnapi_handler.h:20:
In file included from C:/Users/luncl/Documents/vcpkg-registry/buildtrees/tensorflow-lite/src/v2.17.0-6beb5bd9c1/tensorflow/core/platform/logging.h:19:
In file included from C:/Users/luncl/Documents/vcpkg-registry/buildtrees/tensorflow-lite/src/v2.17.0-6beb5bd9c1/tensorflow/core/platform/types.h:22:
In file included from C:/Users/luncl/Documents/vcpkg-registry/buildtrees/tensorflow-lite/src/v2.17.0-6beb5bd9c1/third_party/xla/third_party/tsl/tsl/platform/types.h:22:
In file included from C:/Users/luncl/Documents/vcpkg-registry/buildtrees/tensorflow-lite/src/v2.17.0-6beb5bd9c1/third_party/xla/third_party/tsl/tsl/platform/ml_dtypes.h:19:
C:/Users/luncl/Documents/vcpkg-registry/installed/arm64-android/include/ml_dtypes/float8.h:871:29: error: no type named 'get_integer_by_size' in namespace 'Eigen::numext'
  871 |     typename Eigen::numext::get_integer_by_size<kNumBytes>::unsigned_type;
      |     ~~~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~
C:/Users/luncl/Documents/vcpkg-registry/installed/arm64-android/include/ml_dtypes/float8.h:871:48: error: expected ';' after alias declaration
  871 |     typename Eigen::numext::get_integer_by_size<kNumBytes>::unsigned_type;
      |                                                ^
C:/Users/luncl/Documents/vcpkg-registry/installed/arm64-android/include/ml_dtypes/float8.h:892:20: error: no template named 'GetUnsignedInteger'
  892 |   using BitsType = GetUnsignedInteger<sizeof(Float)>;
      |                    ^