fix(visualize): guard against invalid visual_genre crashing generation

[wangzitiansky]

Description

Problem: Book generation crashes when creating interactive visualization
blocks. The analysis LLM sometimes outputs a visual_genre value outside the
allowed list — for example, "simulation" instead of "interactive":

ValidationError: 1 validation error for VisualizationAnalysis
visual_genre
Input should be '', 'flowchart', 'structural', 'illustrative',
'chart', 'stepper', 'interactive', 'mockup' or 'art'
[type=literal_error, input_value='simulation', input_type=str]

Root cause: The user_template_fixed prompt described visual_genre with
an open-ended phrase ("the genre that best fits..."), so the LLM never saw a
hard boundary on which values are actually allowed. Pydantic's
model_validate then raised a ValidationError with no recovery path,
hard-failing the entire pipeline.

Fix: Two-layer defense:

1. Prompt (prevention): Explicitly enumerate all 8 valid values in
   user_template_fixed (en + zh) so the LLM knows the boundary before
   generating.
2. Code (fallback): Wrap model_validate with try/except in
   AnalysisAgent — on failure, degrade visual_genre to "" and retry.
   The code generator already has an "any other genre" catch-all branch,
   so empty string degrades gracefully instead of crashing.

Related Issues

• Closes #...
• Related to #...

Module(s) Affected

• agents
• api
• config
• core
• knowledge
• logging
• services
• tools
• utils
• web (Frontend)
• docs (Documentation)
• scripts
• tests
• Other: ...

Checklist

• I have read and followed the contribution guidelines.
• My code follows the project's coding standards.
• I have run pre-commit run --all-files and fixed any issues.
• I have added relevant tests for my changes.
• I have updated the documentation (if necessary).
• My changes do not introduce any new security vulnerabilities.

Additional Notes

Add any other context or screenshots about the pull request here.