Improve packaged Linux binary performance

In order to address problems on older CPUs with "illegal instruction"
errors, we set the target CPU to x86-64 for Deb and RPM release builds.
Original issue is #1682 and the PR to implement was #1686.

As part of #1682, we discussed that we should also build using
-mtune=intel to get better performance.

.travis_script.bash

@@ -11,7 +11,7 @@ fi
 
 ponyc-test(){
   echo "Building and testing ponyc..."
-  make CC="$CC1" CXX="$CXX1" test-ci
+  make CC="$CC1" CXX="$CXX1" tune=intel test-ci
 }
 
 ponyc-build-packages(){
@@ -26,7 +26,7 @@ ponyc-build-packages(){
   PACKAGE_ITERATION="${TRAVIS_BUILD_NUMBER}.$(git rev-parse --short --verify 'HEAD^{commit}')"
 
   echo "Building ponyc packages for deployment..."
-  make verbose=1 arch=x86-64 config=release package_name="ponyc" package_base_version="$(cat VERSION)" package_iteration="${PACKAGE_ITERATION}" deploy
+  make CC="$CC1" CXX="$CXX1" verbose=1 arch=x86-64 tune=intel config=release package_name="ponyc" package_base_version="$(cat VERSION)" package_iteration="${PACKAGE_ITERATION}" deploy
 }
 
 ponyc-build-docs(){