Fixing bugs in the baseline and evaluation

[bond005]

1. Incorrect signatures in the examples/calculate_baseline_metrics.ipynb are fixed.
2. The rag-bench package is refactored, and errors in this package are corrected. Some of these errors are:
   • found_ids in the evaluate_rag_results function were processed incorrectly, because they were represented as string instead of an int list;
   • question types were not taken into account when calculating metrics, although they should be according to the evaluator's logic;
   • and so on.
3. In addition to the jupyter notebook, a special python script examples/calculate_baseline_metrics.py is implemented for more convenient launch of the baseline and obtaining metrics.
4. The special readme.md for the rag-bench package is added.

Items 1 and 2 ensure that the benchmark can be used correctly (without the bug fixes described in these items, it is impossible to use the benchmark to evaluate RAG pipelines). Items 3 and 4 simply improve the usability of the benchmark.