XGB gh cut optimization

[GernotMaier]

No description provided.

[Copilot]

Pull request overview

Adds a new workflow to derive and visualize energy/zenith-dependent gamma/hadron separation cuts by maximizing Li & Ma significance using precomputed rate surfaces and the trained classification ROC curves.

Changes:

• Add optimize_classification.py to compute optimized gamma-efficiency cuts (and derived efficiencies/thresholds/significance) and write results to an ECSV table.
• Add plot_optimize_classification.py to generate 1D (per-zenith) and 2D (energy vs zenith) diagnostic plots from the ECSV output.
• Improve classification training/config behavior (fail fast on empty datasets; propagate max_cores/random_state into model hyperparameters) and update the conda environment.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 7 comments.

Show a summary per file

File	Description
src/eventdisplay_ml/scripts/optimize_classification.py	New script to optimize cut surfaces and export results to ECSV.
src/eventdisplay_ml/scripts/plot_optimize_classification.py	New plotting script for inspecting optimization outputs.
src/eventdisplay_ml/models.py	Classification training now errors on empty inputs instead of silently skipping.
src/eventdisplay_ml/config.py	Clarifies --max_events semantics and propagates n_jobs/random_state into model hyperparameters.
environment.yml	Adds astropy-base and pins xgboost.
docs/changes/56.feature.md	Towncrier fragment describing the new cut-optimization feature.