feat(factory): put few known_issues cards into factory and add log

[Dong1017]

Summary

Expand incubating/factory/cards/known_issues with a first batch of high-frequency runtime failure cards so Factory can route common Ascend/GPU failures before deeper reasoning.

What Changed

Added these known_issue cards:

• missing-cann-environment
• device-out-of-memory
• distributed-communication-timeout
• ms-context-empty
• ms-tbe-operator-compilation-error
• stack-version-mismatch

Also updated the Factory manifest inventory in incubating/factory/manifests/pack.yaml so the new cards are included in the stable pack.

Why

The existing known_issues set covered only a small slice of failure signatures. In practice, many first-line diagnosis requests cluster around:

• missing or incomplete CANN setup
• device OOM
• distributed/HCCL timeout
• MindSpore context initialization failures
• TBE compile failures on Ascend
• mixed-version / ABI mismatch after partial upgrades

These are strong candidates for Factory because they are:

• frequent
• easy to detect from stable log/error patterns
• high-value for early routing
• low-ambiguity compared with deeper implementation bugs

Card Design Notes

Each card is kept in the current known_issue schema shape and includes:

• stable id
• symptom: failure
• severity and lifecycle metadata
• platform tags
• detection regex/patterns
• short description
• concise first-line fix summary

The intent is to improve triage and reuse, not to encode full repair playbooks.

Impact

This increases the known_issues inventory in the pack and gives Factory better coverage for bootstrap/runtime failures that appear before model- or operator-specific analysis.

Expected user-facing effect:

• faster routing for common infrastructure/runtime failures
• fewer cases falling through to generic “investigate manually” paths
• cleaner separation between known operational failures and deeper code-level issues

Validation

• Confirm new YAML files conform to the current known_issue card shape.
• Confirm incubating/factory/manifests/pack.yaml includes the new card entries and updated known_issues count.
• Spot-check detection patterns and summaries against the intended runtime signatures.