Fix remote_host_prepare.sh to pass shellcheck (with bugfix)

tools/run_tests/performance/remote_host_prepare.sh

@@ -15,7 +15,7 @@
 
 set -ex
 
-cd $(dirname $0)/../../..
+cd "$(dirname "$0")/../../.."
 
 # TODO(jtattermusch): To be sure there are no running processes that would
 # mess with the results, be rough and reboot the slave here
@@ -24,7 +24,7 @@ ssh "${USER_AT_HOST}" "killall -9 qps_worker dotnet mono node ruby worker || tru
 
 # On Windows, killall is not supported & we need to kill all pending workers
 # before attempting to delete the workspace
-ssh "${USER_AT_HOST}" "ps -e | egrep 'qps_worker|dotnet' | awk '{print $1}' | xargs kill -9 || true"
+ssh "${USER_AT_HOST}" "ps -e | egrep 'qps_worker|dotnet' | awk '{print \$1}' | xargs kill -9 || true"
 
 # cleanup after previous builds
 ssh "${USER_AT_HOST}" "rm -rf ~/performance_workspace && mkdir -p ~/performance_workspace"
@@ -36,4 +36,5 @@ scp ../grpc.tar "${USER_AT_HOST}:~/performance_workspace"
 ssh "${USER_AT_HOST}" "tar -xf ~/performance_workspace/grpc.tar -C ~/performance_workspace || tar -xf ~/performance_workspace/grpc.tar -C ~/performance_workspace"
 
 # For consistency with local run, invoke the kill_workers script remotely.
+# shellcheck disable=SC2088
 ssh "${USER_AT_HOST}" "~/performance_workspace/grpc/tools/run_tests/performance/kill_workers.sh"