Make `EvaluationFramework` work

[kuraisle]

Is this the right issue type?

• [x] Yes, I'm planning work for this project team.

Summary

Currently the EvaluationFramework class doesn't do anything. We need to make it useful.

The behaviour we need is:

• The class loads data from a file. The data will contain the input and expected output
• The class will initialise with a list of PipelineTests
• run_evaluations should run each input through each PipelineTest and evaluates the output
  • I would suggest all inputs are run through a TestPipeline and then the pipeline is dropped for RAM purposes, particularly with LLM pipelines
• After running, _save_evaluations() should save the results to an output file

Acceptance Criteria

• [ ] EvaluationFramework has data loaders
• [ ] A list of PipelineTest added to EvaluationFramework.__init__()
• [ ] EvaluationFramework.run_evaluations() runs input
• [ ] EvaluationFramework.run_evaluations() evaluates output
• [ ] EvaluationFramework._save_evaluations() saves json of the output of running evaluation
• [ ] Add a drop method to pipelines to accommodate dropping LLMPipelines
• [ ] Tests pass.
• [ ] Documentation added.

Tasks

• [ ] Update parameters
• [ ] write methods
• [ ] Unit tests.
• [ ] Documentation.

Confirm creation

• [x] This issue is ready