[BUG] Profiler tool does not catch the RAPIDS jars correctly

[amahussein]

Describe the bug

Currently the CollectInformation.scala checks for the environment details map "Classpath Entries" to verify that RAPIDS jars are included in the classPath.

I found out that this is not always the case. For example, some GPU eventlogs do not have the RAPIDS jars as part of the "Classpath Entries". Instead:

{
   "Event":"SparkListenerEnvironmentUpdate",
   "JVM Information":{
      "Java Home":"/usr/lib/jvm/temurin-8-jdk-amd64/jre",
      "Java Version":"1.8.0_345 (Temurin)",
      "Scala Version":"version 2.12.14"
   },
  "Spark Properties":{
      .....
      "spark.yarn.dist.jars":"file:///tmp/232c28608dc64730b2c0f0cefaa32a05/rapids-4-spark_2.12-22.10.0.jar",
      "spark.yarn.secondary.jars":"rapids-4-spark_2.12-22.10.0.jar",
      "spark.repl.local.jars":"file:///tmp/232c28608dc64730b2c0f0cefaa32a05/rapids-4-spark_2.12-22.10.0.jar",
  },
   "System Properties":{
   "sun.java.command":"org.apache.spark.deploy.SparkSubmit --conf spark.executor.memory=16G  --jars /tmp/232c28608dc64730b2c0f0cefaa32a05/rapids-4-spark_2.12-22.10.0.jar /tmp/232c28608dc64730b2c0f0cefaa32a05/sample_sanity.py gs://spark-directory",
  }

Some other properties to hold the RAPIDS jars:

spark.driver.extraClassPath
spark.executor.extraClassPath

This means that profiler will pass wrong information to the AutoTuner.

• AutoTuner will mistakenly considered the path as loaded.
• AutTuner won't make the recommendation based on the jars version