feat(markdown): font-independent Unicode glyph mapping

[frederikbeimgraben]

Summary

Replaces the single hard-coded euro fix in the Markdown converter with a
data-driven table mapping non-renderable Unicode characters to
font-independent TeX nodes. tectonic (XeTeX) does no font fallback, so a code
point the bundled DIN text font lacks would otherwise render as a blank "tofu"
box.

Mapping table (pytex_markdown/glyphs.py)

Char	Target	Package
€	\euro{}	eurosym
→	$\rightarrow$	— (base TeX)
↔	$\leftrightarrow$	—
≤	$\leq$	—
≥	$\geq$	—
·	$\cdot$	—

• The arrow targets match the existing ASCII-arrow rewrites, so → and ->
  typeset identically.
• · maps to the math \cdot (multiplication dot) rather than the
  font-dependent \textperiodcentered, which would itself tofu under DIN.
• _prose() now splits prose generically over the table (functional, genexp
  style); the euro is the first entry, no longer a special case. Code
  spans/blocks stay verbatim.

Trust levels

eurosym added to the UNTRUSTED/SANDBOXED package allowlist (_policy.py),
so a € in untrusted Markdown renders instead of being rejected with a
TrustError. The math targets pull no package.

Missing-glyph handling

A character that is neither mapped nor present in every bundled DIN
weight is genuinely unrenderable → replaced by a \texttt{[missing glyph]}
placeholder and a MissingGlyphWarning naming the char + U+XXXX, instead
of silent tofu. DIN coverage is read by a zero-dependency cmap parser over the
bundled fonts; the rule is conservative (renderable only if present in every
weight).

Tests

• Per-char mapping (€ → ↔ ≤ ≥ ·).
• UNTRUSTED and SANDBOXED render of all mapped chars — no TrustError,
  no network/compile.
• Missing glyph → [missing glyph] placeholder + warning.
• Euro regression (existing test_euro.py stays green).
• Code spans untouched; renderable chars (umlauts, dashes) pass through.

Status

• Full suite: 859 passed, 2 skipped.
• basedpyright src: 0 errors, 0 warnings from changed files.
• ruff format --check + ruff check: clean.

Wiki updated (Markdown-to-PDF "Unicode glyph handling" section, Blob-API
allowlist note).

🤖 Generated with Claude Code