fix ray benchmark

[anson627]

This pull request updates the Ray job controller configuration and templates to improve parameterization and flexibility, especially for image and resource specifications. The main changes include making image and GPU resource values configurable via template variables, fixing a typo in a configuration key, and improving the way configuration paths are resolved in the Python code.

Template and Configuration Improvements:

• Updated job_template.yaml to use template variables for the container image ({{.Image}}) and GPU resource requests/limits ({{.JobGPU}}) for ray-submitter, ray-head, and ray-worker containers, allowing dynamic substitution during job creation. [1] [2] [3]
• Changed objectTemplatePath in config.yaml to use a template variable ({{$job_template_path}}), making the job template path configurable.

Bug Fixes and Path Handling:

• Fixed a typo in config.yaml: corrected measurmentInterval to measurementInterval.
• Updated the install_ray_dependencies method in job_controller.py to use the instance variable self.cl2_config_dir when resolving the Ray config directory, improving flexibility for different environments.

[Copilot]

Pull request overview

This PR fixes a typo in the Ray benchmark configuration file. The parameter name measurmentInterval (missing an 'e') has been corrected to measurementInterval in the SchedulingThroughput measurement configuration.

Changes:

• Corrected spelling of configuration parameter from measurmentInterval to measurementInterval