Update outdated dependency versions in packages.txt for Python 3.12 compatibility

[takuan517]

Summary

src/configs/packages.txt specifies dependency versions from 2020, which are increasingly incompatible with modern Python environments such as Ubuntu 24.04 (Python 3.12).

Current State

scipy==1.5.2        # Released 2020
pandas==1.1.3       # Released 2020
numpy==1.19.2       # Released 2020
torch==1.4.0        # Released 2020
xgboost==0.90       # Released 2019
matplotlib==3.3.2   # Released 2020

Known Breakage

pandas==1.1.3 is the most critical case. The .max()[0] pattern used across 51 table_generator.py files silently works on pandas v1.x but raises KeyError: 0 on pandas v2.0+. This has been fixed in PR #7, but the root cause is the outdated dependency specification.

Other packages likely have similar silent API changes that have not yet surfaced.

Proposed Work

• Audit each pinned dependency for API-breaking changes introduced
  since the specified version
• Update packages.txt and requirements_pip3.txt to versions
  compatible with Python 3.12
• Test the full Planter workflow (data loading → model training →
  table generation → BMv2 deployment) against the updated dependencies
• Document the tested environment (Python version, OS, package versions)
  in the README or a dedicated COMPATIBILITY.md

Environment Context

Modern default environments that trigger breakage:

• Ubuntu 24.04 LTS: Python 3.12, pandas 2.x, numpy 1.26+
• macOS 14+ with Homebrew: Python 3.12, pandas 2.x

Related

• PR fix: replace .max()[0] with .max().iloc[0] for pandas v2.x compatibility #7: immediate fix for the pandas .max()[0] breakage