kubernetes local should work with in process launcher (#601)

## Upstream SPARK-XXXXX ticket and PR link (if not applicable, explain)

local kubernetes submission is not submitted upstream yet

## What changes were proposed in this pull request?

In process launcher clobbers spark configuration with launcher port and secret overrides, resulting the local spark submit on the driver to connect to the port used by the cluster mode spark submit process.

Spark drivers in k8s are hard coded to run spark-submit, so spark launcher overrides are inapplicable to them. Excluding the configurations from the config map properly signals the local backend to not try to connect to a launcher server

resource-managers/kubernetes/core/src/main/scala/org/apache/spark/deploy/k8s/submit/KubernetesClientApplication.scala

@@ -176,7 +176,9 @@ private[spark] class Client(
   // Build a Config Map that will house spark conf properties in a single file for spark-submit
   private def buildConfigMap(configMapName: String, conf: Map[String, String]): ConfigMap = {
     val properties = new Properties()
-    conf.foreach { case (k, v) =>
+    conf
+      .filter { case (k, _) => !k.startsWith("spark.launcher")}
+      .foreach { case (k, v) =>
       properties.setProperty(k, v)
     }
     val propertiesWriter = new StringWriter()