Figure3-4-5 Naive/IGS/FGS performance fgs

[zacharopoulou]

Summary

This PR adds manuscript figure generation for performance evaluations across three miRNA-gene pair universes:

Figure	Evaluation	Universe of miRNA-gene pairs
Figure 3	Naive / algorithm-specific evaluation	Each algorithm is evaluated only on the ground-truth pairs for
which that algorithm supplied a score. Therefore, every algorithm may be evaluated on a different gene/pair set.
Figure 4	IPS - Intersection Pair Set	Only ground-truth miRNA-gene pairs for which **all evaluated
algorithms supplied a prediction score**. This is one shared, restricted universe used for every algorithm.
Figure 5	FPS - Full Pair Set	Every miRNA-gene pair in the complete benchmark ground-truth dataset. When an
algorithm did not score a pair, it receives the defined zero-equivalent score.

The branch keeps the generated figures and tables out of git and documents the commands needed to regenerate them.

Changes

• Adds separate figure entry points:
  • scripts/figure3_performance_naive.py
  • scripts/figure4_performance_ips.py
  • scripts/figure5_performance_fps.py
• Adds shared performance plotting/statistics helpers:
  • scripts/figure_performance_common.py
• Generates each panel separately before assembling combined figures.
• Uses box plots with per-dataset dots for metric distributions.
• Adds APS and Spearman R2 leaderboard panels.
• Keeps random baseline in Figures 4 and 5.