Reversed the order of tests to find a scala executable (in the case when...

Reversed the order of tests to find a scala executable (in the case when SPARK_LAUNCH_WITH_SCALA is defined): instead of checking in the PATH first, and only then (if not found) for SCALA_HOME, now we check for SCALA_HOME first, and only then (if not defined) do we look in the PATH.  The advantage is that now if the user has a more recent (non-compatible) version of scala in her PATH, she can use SCALA_HOME to point to the older (compatible) version for use with spark.

Suggested by Josh Rosen in this thread:

  https://groups.google.com/forum/?fromgroups=#!topic/spark-users/NC9JKvP8808

run

@@ -47,14 +47,15 @@ case "$1" in
 esac
 
 if [ "$SPARK_LAUNCH_WITH_SCALA" == "1" ]; then
-  if [ `command -v scala` ]; then
-    RUNNER="scala"
+  if [ "$SCALA_HOME" ]; then
+    RUNNER="${SCALA_HOME}/bin/scala"
   else
-    if [ -z "$SCALA_HOME" ]; then
-      echo "SCALA_HOME is not set" >&2
+    if [ `command -v scala` ]; then
+      RUNNER="scala"
+    else
+      echo "SCALA_HOME is not set and scala is not in PATH" >&2
       exit 1
     fi
-    RUNNER="${SCALA_HOME}/bin/scala"
   fi
 else
   if [ `command -v java` ]; then