fix(renderer): honor nested fence lengths

[Michaelyklam]

Thinking Path

• Hermes WebUI's renderer already line-anchors fenced blocks to avoid bug(renderer): triple backticks mid-line treated as fence delimiters — corrupts downstream markdown (eats *, leaks </strong>) — fence regex needs CommonMark line anchoring #1438-style mid-line fence corruption.
• CommonMark §4.5 also requires the closing fence to use the same character and at least as many backticks as the opener.
• The current regexes still hard-coded triple-backtick closers, so 4/5-backtick markdown examples closed at inner triple fences.
• This PR updates the renderer paths to track the opener length, preserves language-header behavior, and pins the regression against the actual renderMd() JS driver plus the Python mirror.

What Changed

• Updated static/ui.js fenced-code matching to capture {3,} backtick opener runs and require \2 plus optional extra backticks on close.
• Applied the same fence-length rule to user-message fenced rendering and to the blockquote pre-pass fence-state walker.
• Kept normal 3-backtick language rendering and empty-fence handling covered.
• Updated source/driver tests that extract renderMd() so they include the new shared fence helpers.

Why It Matters

• Fixes nested markdown/code examples where an outer 4+ or 5+ backtick fence contains inner triple-backtick fences.
• Prevents stray outer closers from leaking into prose and keeps the code block as one intact <pre><code> block.

Fixes #1696

Verification

• pytest tests/test_sprint16.py::test_render_md_fenced_code_with_five_backtick_outer_preserves_inner_triples tests/test_sprint16.py::test_render_md_fenced_code_with_four_backtick_outer_preserves_inner_triples tests/test_sprint16.py::test_render_md_fenced_code_three_backtick_path_still_renders_language tests/test_renderer_js_behaviour.py::TestFencedCodeFenceLength tests/test_1325_user_fenced_code.py::TestUserFencedBlocks::test_four_backtick_outer_fence_preserves_inner_triple_fence -q — 7 passed
• pytest tests/test_sprint16.py tests/test_renderer_js_behaviour.py tests/test_1325_user_fenced_code.py tests/test_issue1438_fence_anchoring.py tests/test_issue1154_fenced_code_leak.py tests/test_issue1618_yaml_json_diff_newline_preserve.py tests/test_issue1446_glued_heading_lift.py tests/test_issue_code_syntax_highlight.py -q — 197 passed
• git diff --check — passed
• env -u HERMES_CONFIG_PATH /home/michael/.hermes/hermes-agent/venv/bin/python -m pytest tests/ -q — 4484 passed, 2 skipped, 3 xpassed, 1 warning, 8 subtests passed in 407.20s

No screenshots included: this is a renderer parsing fix verified at the rendered-HTML level with the actual static/ui.js renderMd() node driver.

Risks / Follow-ups

• This keeps the existing backtick-only scope; tilde fences were not added.
• Info strings that are not a single language token still do not produce a language header, matching the existing conservative language handling.

Model Used

• OpenAI Codex gpt-5.5 via Hermes Agent CLI.
• Tool use: Hermes file tools, shell/git/pytest, GitHub CLI.

[nesquena-hermes]

Closed by the v0.51.4 release in PR #1707 (merged at 4daa238, deployed to production).

Live on production: https://github.com/nesquena/hermes-webui/releases/tag/v0.51.4

🚀