[RAGAS] fix: Metric parameter validation and metric descriptors

[julian-risch]

fixes #528

• Extends metric param validation, so that it is done by our integration framework and not Ragas

[julian-risch]

@shadeMe now we don't distinguish required and optional parameters. name is used only for ASPECT_CRITIQUE, where it is needed.

Note that the implementation of def _init_metric(self): in deepeval and now in ragas checks that all metric init parameters provided by the user are in the list of expected metric init parameters. In addition, it checks that if the metric expects any init parameters then the user must provide at least one of those. It doesn’t check that all of them are provided by the user. So it’s not fully checking required parameters but also not checking only allowed parameters.

Here is the part of code that I am referring to: https://github.com/deepset-ai/haystack-core-integrations/pull/555/files#diff-c0c023551bfa22d8d23e1e3f6cefcae07b3819de9b9811e313126f6bae1c5a53R81