[UI][EDP] Schema for "args" is incorrect for Pig jobs

For Pig jobs, job_configs["args"] should be a list of strings. Savanna currently requires it to be a dictionary and generates a workflow which is incorrect but will still run.

Here is some background:

Oozie allows <param> tags and <argument> tags in pig actions. Both are used to pass options to the Pig script (Pig usage is shown at the end of this description).

For Oozie pig actions

  <param>name=value</param>

is short hand for

  <argument>-param</argument>
  <argument>name=value</argument>

So, job_configs["params"] should be a dictionary (as it is) but job_configs["args"] should be a list of literal strings which are passed as Pig options, separated by spaces.

Savanna currently uses <param> tags to set the INPUT and OUTPUT values for the pig script based on data sources.
Job_configs["args"] would be used to set additional flags (see the Pig usage below).

As an example, the following 2 Oozie workflows are equivalent:

<?xml version="1.0" ?>
<workflow-app name="job-wf" xmlns="uri:oozie:workflow:0.2">
  <start to="job-node"/>
  <action name="job-node">
    <pig>
      <job-tracker>${jobTracker}</job-tracker>
      <name-node>${nameNode}</name-node>
      <configuration>
        <property>
          <name>fs.swift.service.savanna.password</name>
          <value>password</value>
        </property>
        <property>
          <name>fs.swift.service.savanna.username</name>
          <value>user</value>
        </property>
        <property>
           <name>oozie.use.system.libpath</name>
           <value>true</value>
        </property>
      </configuration>
      <script>pig.script</script>
      <param>INPUT=swift://tmckay.savanna/input</param>
      <param>OUTPUT=swift://tmckay.savanna/output</param>
    </pig>
    <ok to="end"/>
    <error to="fail"/>
  </action>

<?xml version="1.0" ?>
<workflow-app name="job-wf" xmlns="uri:oozie:workflow:0.2">
  <start to="job-node"/>
  <action name="job-node">
    <pig>
      <job-tracker>${jobTracker}</job-tracker>
      <name-node>${nameNode}</name-node>
      <configuration>
        <property>
          <name>fs.swift.service.savanna.password</name>
          <value>password</value>
        </property>
        <property>
          <name>fs.swift.service.savanna.username</name>
          <value>user</value>
        </property>
      </configuration>
      <script>pig.script</script>
      <argument>-param</argument>
      <argument>OUTPUT=swift://tmckay.savanna/output</argument>
      <argument>-param</argument>
      <argument>INPUT=swift://tmckay.savanna/input</argument>
    </pig>
    <ok to="end"/>

Here is Pig usage, taken from Oozie logs on a failed job:

Apache Pig version 0.10.1 (r1426282)
compiled Dec 27 2012, 11:24:26

USAGE: Pig [options] [-] : Run interactively in grunt shell.
       Pig [options] -e[xecute] cmd [cmd ...] : Run cmd(s).
       Pig [options] [-f[ile]] file : Run cmds found in file.
  options include:
    -4, -log4jconf - Log4j configuration file, overrides log conf
    -b, -brief - Brief logging (no timestamps)
    -c, -check - Syntax check
    -d, -debug - Debug level, INFO is default
    -e, -execute - Commands to execute (within quotes)
    -f, -file - Path to the script to execute
    -g, -embedded - ScriptEngine classname or keyword for the ScriptEngine
    -h, -help - Display this message. You can specify topic to get help for that topic.
        properties is the only topic currently supported: -h properties.
    -i, -version - Display version information
    -l, -logfile - Path to client side log file; default is current working directory.
    -m, -param_file - Path to the parameter file
    -p, -param - Key value pair of the form param=val
    -r, -dryrun - Produces script with substituted parameters. Script is not executed.
    -t, -optimizer_off - Turn optimizations off. The following values are supported:
            SplitFilter - Split filter conditions
            PushUpFilter - Filter as early as possible
            MergeFilter - Merge filter conditions
            PushDownForeachFlatten - Join or explode as late as possible
            LimitOptimizer - Limit as early as possible
            ColumnMapKeyPrune - Remove unused data
            AddForEach - Add ForEach to remove unneeded columns
            MergeForEach - Merge adjacent ForEach
            GroupByConstParallelSetter - Force parallel 1 for "group all" statement
            All - Disable all optimizations
        All optimizations listed here are enabled by default. Optimization values are case insensitive.
    -v, -verbose - Print all error messages to screen
    -w, -warning - Turn warning logging on; also turns warning aggregation off
    -x, -exectype - Set execution mode: local|mapreduce, default is mapreduce.
    -F, -stop_on_failure - Aborts execution on the first failed job; default is off
    -M, -no_multiquery - Turn multiquery optimization off; default is on
    -P, -propertyFile - Path to property file