Prefer ≟ and add equality aliases

[isomorphisms]

Prefer ≟ as IR's user-facing equality spelling and accept ?=, =?, ?=?, and ¿=? as aliases, alongside existing = and ==.

The new spellings canonicalize to the existing == operation, preserving equality precedence and behavior. Single = remains contextual for named arguments and formal defaults.

Lexer invariant: IR intentionally retains its broad user-definable Unicode infix fallback. Reserved glyphs are recognized first by explicit lexer rules; only unreserved glyphs fall through to generic SPECIAL handling. In particular, ≟ becomes EQ, not SPECIAL, while an unreserved glyph such as ⊗ remains usable as an ordinary infix function.

code/tests/ir-unicode-lexer.R exercises that boundary. Every hexadecimal code point in that regression has the rendered glyph on the same line, and the test checks the mapping itself (including 0x03c0 → π) so those comments cannot silently disagree with what intToUtf8 prints. It then defines and calls a function literally named π, checks the four assignment arrows and all requested equality aliases, and defines/evaluates ⊗ through the generic infix path.

Current head 909cb9db888d5c666c3f0bc5cba7ee84d5d14fe0 has a successful Ubuntu 24.04 GitHub Actions build: dependencies, configure, full build, and the focused Unicode lexer regression all pass. Both the pull-request and same-head push runs completed successfully.

Closes #4.