docs: make the roadmap actionable

[LauraGPT]

Summary

• turn the bilingual repository roadmap into an actionable contributor queue with live good first issue, help wanted, and ready for PR searches
• document label semantics, claim-before-coding workflow, acceptance evidence, and issue/PR closure boundaries
• identify current hardware and upstream-review needs for realtime L20 profiling, AMD Windows Vulkan, checkpoint publication, and Transformers integration
• update the current release boundary to FunASR 1.4.11 and clarify that MOSS emits anonymous speaker labels, not known-person identities
• refresh 1.4.3 的 funasr/bin/realtime_ws.py 的 run_session_work() 实现太简单粗暴，导致并发性能比1.3.9还差很多 #3528 as an equal-work preview-efficiency investigation and add the tested Qwen3-ASR vLLM workflow under review

Related to #3203 and #3528. This PR does not close either issue; both still require community or reporter validation.

Verification

• python -m pytest -q tests/test_docs_funasr_install_commands.py (18 passed)
• bilingual roadmap invariants and local relative Markdown links verified
• git diff --check passed
• GitHub release v1.4.11 and PyPI latest 1.4.11 independently verified before updating the version boundary

Review focus

• contributor tasks are phrased as acceptance evidence rather than promised delivery dates
• live label queries remain useful as issue state changes
• English and Chinese pages carry equivalent scope and closure policy