[Repo Assist] Encode emoji/high-Unicode as HTML entities in all HTML output paths

[github-actions]

🤖 This is an automated PR from Repo Assist, implementing the code review suggestions from PR #965.

Addresses the three unresolved review threads in #965.

What this PR does

Implements the code review suggestions from PR #965:

1. Consistent emoji encoding across all HTML output paths (Thread 3)

PR #965 added encodeHighUnicode but only applied it to Literal spans. This PR applies it inside htmlEncode itself so that InlineCode, CodeBlock, attribute values (alt, title), and all other paths automatically encode emoji/high-Unicode as numeric HTML entities — not just Literal spans.

Literal spans (which bypass htmlEncode) also get encodeHighUnicode applied directly.

2. Fix misleading comment (Thread 4)

The original comment said "Encode all characters outside BMP (>= 0x10000) as they're typically emojis". Rewrote to clarify the actual rationale: encoding non-BMP code points and selected BMP ranges to avoid output encoding issues, since many non-emoji scripts also live outside the BMP.

3. Variation selector encoding + test assertion (Thread 5)

The ⚠️ sequence consists of ⚠ (U+26A0) followed by variation selector VS16 (U+FE0F). The review noted that the test checked for ⚠ (⚠) but not for ️ (VS16). This PR:

• Extends the encoded BMP ranges to include variation selectors U+FE00–U+FE0F
• Adds the assertion html |> should contain "️" to the emoji test

Note: The PR also narrows the broad U+2000–U+2BFF range from #965 to targeted emoji/symbol blocks only (U+2600–U+26FF, U+2700–U+27BF, U+2B00–U+2BFF). This avoids breaking CommonMark spec tests that expect box-drawing characters (e.g., U+255C ╜) to pass through unchanged in <pre><code> blocks.

Tests

All 182 markdown tests pass, including the CommonMark spec suite.

New tests added:

• Emojis are encoded as HTML numeric entities — verifies 🎉🚧⭐⚠️✅ all encode correctly, including the variation selector
• Regular text without emojis is not modified — verifies Cyrillic and Chinese characters are not touched
• List without blank line after heading — covers the Conversions from fsx comments markdown to html containing emoji (on AI-generated documents) #964 scenario

Test Status

✅ dotnet test tests/FSharp.Markdown.Tests/FSharp.Markdown.Tests.fsproj — 182 passed, 0 failed

Generated by Repo Assist

To install this workflow, run gh aw add githubnext/agentics/workflows/repo-assist.md@828ac109efb43990f59475cbfce90ede5546586c. View source at https://github.com/githubnext/agentics/tree/828ac109efb43990f59475cbfce90ede5546586c/workflows/repo-assist.md.

[github-actions]

✅ Pull request created: #986

[nojaf]

@dsyme as this is in a draft state, what should I make out of this?
Does this need a review? Is this unfinished?
It is a little unclear from afar who is supposed to make the next move here.

[dsyme]

@nojaf This PR was the result of me asking Repo Assist to fix up #965. This comment

#965 (comment)

It's not indicative of what Repo Assist would normally do autonomously (it would usually start with an issue if at all, and never work on something so obscure)

It also incorrectly made a new PR targeting main, effectively a replacement PR, instead of fixing up the PR itself. I'll fix that for future.

Next step is I'll manually merge this into OG PR #965 and we can decide what to do there.

[dsyme]

Oh I see, the OG PR came from a fork so Repo Assist couldn't fix it up.