refactor(examples): Consolidate and clean up example scripts

[m96-chan]

Problem

Multiple large example files with overlapping functionality:

• examples/chat_cli.py (990 lines)
• examples/chat_cli_thinking.py (714 lines)
• examples/chat_cli_moe.py (572 lines)
• examples/demo_v02_full.py (633 lines)
• examples/demo_v0210.py (593 lines)

Issues

• Duplicated code across examples
• Version-specific demos become stale
• Hard to maintain

Proposed Structure

examples/
├── chat/
│   ├── basic.py         (simple chat)
│   ├── streaming.py     (streaming output)
│   ├── thinking.py      (thinking mode)
│   └── moe.py           (MoE models)
├── inference/
│   ├── benchmark.py     (performance testing)
│   └── batch.py         (batch inference)
├── audio/
│   ├── stt.py           (speech-to-text)
│   └── realtime.py      (real-time processing)
└── common/
    └── utils.py         (shared utilities)

Benefits

• No duplicated code
• Clear categorization
• Easier to maintain
• Remove stale version-specific demos

[m96-chan]

Completed in PR #168