UPSTREAM PR #17007: webui: fix keyboard shortcuts for new chat & edit chat title

[DajanaV]

Mirrored from ggml-org/llama.cpp#17007

As per spec: https://www.w3.org/TR/uievents-key/#example-739eafdc

Built and tested. Static build included.

[loci-review]

Access the complete analysis in the LOCI Dashboard

Performance Analysis Summary

Based on the comprehensive analysis of project_id=2621b8c0-b5ce-11f0-b333-453f42058aa1 comparing versions ba4ca076-88f5-47e7-bb43-7f8057f51cee against 523c96f3-1c47-4c5d-b466-d0a3e39ed76e, the changes show minimal performance impact with no modifications to core LLM inference functions.

Key Findings

Performance Metrics:

• Highest Response Time change: linenoiseBeep function (+0.17%, 76 ns → 76 ns)
• Highest Throughput change: linenoiseBeep function (+0.21%, 61 ns → 61 ns)
• Changes are sub-nanosecond level and within measurement noise

Core Function Impact:
No changes detected in critical inference functions:

• llama_decode(), llama_encode(), llama_tokenize() remain unmodified
• Model loading, memory management, and batch processing functions unchanged
• Tokens per second impact: None - no inference-critical functions affected

Power Consumption Analysis:
Negligible changes across all binaries (< 0.01%):

• build.bin.libllama.so: -0.0002% decrease
• build.bin.llama-tts: -0.0001% decrease
• All other binaries show zero measurable change

Flame Graph & CFG Analysis:

• linenoiseBeep shows simple execution structure with 81.3% self-time
• Identical assembly code between versions confirms no functional changes
• Performance variations attributed to measurement noise and system-level factors

GitHub Code Review:
Analysis Discrepancy Identified: PR #82 contains only WebUI keyboard shortcut fixes (changing 'o'/'e' to 'O'/'E' for Shift key compliance), but performance analysis measured linenoiseBeep function. The performance measurements appear unrelated to the actual code changes in the pull request.

Conclusion:
The performance analysis shows stable execution with variations within measurement precision. The actual code changes in PR #82 affect only frontend JavaScript event handling with no impact on LLM inference performance. The linenoiseBeep timing variations require separate investigation as they don't correspond to the WebUI changes.