Model Filtering: Extend to multiple benchmarks

[vijusudhi]

Currently, models are filtered based on the results from OpenLLM Benchmark. However this is already archived and do not support new models.

TODOs:

• Extend to other recent benchmarks
  For example: https://artificialanalysis.ai/leaderboards/models
• For each benchmark, define a mapper - from the platform tasks to the different datasets in the benchmark
• More filters including:
  - inference time
  - language
  - co2 emissions, etc.