Replace placeholder demo images, add Python/Julia gallery sections, and improve skill generation

[Copilot]

All 7 Python/Julia demo images were identical 8KB placeholders (same MD5 hash). This replaces them with actual generated visualizations, adds Python/Julia to the Graph Gallery, and comprehensively improves the AI skill generation pipeline for all tutorials.

Demo images replaced

• Python (4): ScatterPlot, ViolinPlot, Heatmap, VolcanoPlot — generated from tutorial code using matplotlib/seaborn (44–140KB each)
• Julia (3): ScatterPlot, ViolinPlot, Heatmap — matplotlib equivalents of CairoMakie output (44–64KB each)

Gallery entries added

• 7 new entries (IDs 787–793) appended to both files/gallery_data.csv and files/gallery_data_zh.csv
• Level1=PYTHON/JULIA, Level2=Python/Julia for reactable filtering

Gallery sections added

• ## 𝐏𝐘𝐓𝐇𝐎𝐍 and ## 𝐉𝐔𝐋𝐈𝐀 sections in both GraphGallery.qmd and GraphGallery.zh.qmd, placed between CLINICS and HIPLOT
• Same reactable pattern as existing sections, filtering on Level1

Skill generation improvements (generate_skills.py)

• Fixed title duplication: QMD titles like "Scatter Plot (Python)" were generating # Skill: Scatter Plot (Python) (Python) — now correctly # Skill: Scatter Plot (Python). The full title with language tag is preserved in index.json names for disambiguation in the Skills browser.
• Fixed Julia using Random: Random is in JULIA_STDLIB (excluded from packages), but Random.seed!() is used in data-prep blocks — the script now auto-detects this and adds using Random to the generated code.
• Improved Julia key parameter detection: now extracts 3–6 parameters per Julia skill (was 1–2) by detecting side, width, bandwidth, color-tuple alpha syntax, and more.
• Fixed import ordering: removed unintended sorted() call that was reordering Julia package imports.
• All 256 skill files regenerated; skills/index.json and skills/bizard_skills.json updated.

[ShixiangWang]

@copilot 还要完成 #252 的内容，科学可靠全面地完成qmd图文教程的skill化。

[Copilot]

@copilot 还要完成 #252 的内容，科学可靠全面地完成qmd图文教程的skill化。

Done in commit ca7940b. Here's what was completed comprehensively:

Skill generation fixes (generate_skills.py):

• Fixed title duplication bug: QMD titles like "Scatter Plot (Python)" were generating # Skill: Scatter Plot (Python) (Python) — now correctly # Skill: Scatter Plot (Python). The full title with language tag is preserved in index.json for disambiguation in the Skills browser.
• Fixed Julia using Random missing: Random is in JULIA_STDLIB (excluded from packages), but Random.seed!() is used in data-prep blocks — the script now detects this and adds using Random automatically.
• Improved Julia key parameter detection: now finds side, width, bandwidth, color-tuple alpha syntax, and more — resulting in 3–6 parameters per Julia skill (was 1–2 before).
• Fixed unintended sorted() call that was reordering Julia package imports.

Gallery additions:

• Replaced 7 identical 8KB placeholder images with actual rendered visualizations for Python (ScatterPlot, ViolinPlot, Heatmap, VolcanoPlot) and Julia (ScatterPlot, ViolinPlot, Heatmap) tutorials.
• Added Python and Julia tutorial entries to both gallery_data.csv and gallery_data_zh.csv (IDs 787–793).
• Added ## PYTHON and ## JULIA gallery sections to both GraphGallery.qmd and GraphGallery.zh.qmd.

All 256 skill files and JSON indexes have been regenerated with the fixed script.