chore(demo-web): add MiniMax M3 as default model

[kimhongyeon]

Affected Package(s)

• @heripo/pdf-parser
• @heripo/document-processor
• @heripo/model
• @heripo/logger
• @heripo/shared (internal)
• demo-web
• docs
• Other (root configs, CI, etc.)

Summary

This PR updates the demo web app to expose MiniMax M3 as an available Together-hosted LLM and uses it for several default processing tasks. The change keeps the upload workflow aligned with the current preferred model set while preserving pricing visibility for cost estimates.

Changes

• Added together/MiniMaxAI/MiniMax-M3 to LLM_MODELS in llm-models.ts with the MiniMax M3 label and Together provider.
• Replaced several default openai/gpt-5.4-mini and LM Studio model references in DEFAULT_FORM_VALUES with together/MiniMaxAI/MiniMax-M3.
• Updated defaults for documentValidationModel, tableCorrection, reviewAssistanceTasks.tables, tocExtractorModel, and visionTocExtractorModel.
• Added MiniMaxAI/MiniMax-M3 pricing metadata to MODEL_PRICING with the same input/output rates as MiniMax M2.5.

Breaking Changes

• None
• If any, describe migration steps:

Checklist

• I followed the Code of Conduct (see CODE_OF_CONDUCT.md)
• I ran: pnpm lint pnpm typecheck pnpm build pnpm test
• Tests added/updated; coverage remains 100%
• Docs/README updated if needed
• No external side effects (network/file/process/time) in tests — use mocks

Related Issues

Closes #