Use gcc-8 instead of just gcc (which may be 9 on arch)

Author: FloopCZ

Dockerfiles/install-archlinux.sh

@@ -30,7 +30,6 @@ pacman -Syu --noconfirm --needed \
 
 if $shared; then
     pacman -S --noconfirm --needed \
-      gcc \
       java-environment=8 \
       libarchive \
       protobuf \

tensorflow_cc/cmake/build_tensorflow.sh.in

@@ -67,8 +67,8 @@ if [ "@ALLOW_CUDA@" = "ON" ] && [ -n "${CUDA_TOOLKIT_PATH}" ]; then
 
     # choose the right version of CUDA compiler
     if [ -z "$GCC_HOST_COMPILER_PATH" ]; then
-        if   hash gcc 2>/dev/null && version_gt 8.5 `gcc -dumpversion`; then
-            export GCC_HOST_COMPILER_PATH=${GCC_HOST_COMPILER_PATH:-"/usr/bin/gcc"}
+        if   hash gcc-8 2>/dev/null && version_gt 8.5 `gcc-8 -dumpversion`; then
+            export GCC_HOST_COMPILER_PATH=${GCC_HOST_COMPILER_PATH:-"/usr/bin/gcc-8"}
         elif hash gcc-7 2>/dev/null && version_gt 7.5 `gcc-7 -dumpversion`; then
             export GCC_HOST_COMPILER_PATH=${GCC_HOST_COMPILER_PATH:-"/usr/bin/gcc-7"}
         elif hash gcc-6 2>/dev/null && version_gt 6.4 `gcc-6 -dumpversion`; then