tamer: nir::interp: Ignore Text tokens

The documentation explains the intent here---existing LaTeX documentation.

The intent was to simply copy the documentation into a LaTeX document based
on the lvspec package that I had created long ago.  Of course, that's not
appropriate---we're a DSL and should provide first-class support for
documentation that will compile properly into the target format, whether it
be LaTeX, HTML, JS, or anything else.

DEV-13162

tamer/src/nir/interp.rs

@@ -250,6 +250,13 @@ impl ParseState for InterpState {
         let span = tok.span();
 
         match self {
+            // TODO: For now we must maintain BC with existing documentation
+            //   that was written as LaTeX,
+            //     which naturally contains many curly braces.
+            // This can go away once we introduce a proper documentation
+            //   language.
+            Ready if matches!(tok, Nir::Text(..)) => Transition(Ready).ok(tok),
+
             // When receiving a new symbol,
             //   we must make a quick determination as to whether it
             //   requires desugaring.

tamer/src/nir/interp/test.rs

@@ -63,6 +63,22 @@ fn does_not_desugar_tokens_without_symbols() {
     );
 }
 
+// We ought to desugar text at some point,
+//   but we need to maintain BC with existing systems.
+// The literate documentation was intended to be LaTeX back in the day,
+//   rather than a proper abstraction atop of it,
+//   and so there are many curly braces in existing code.
+#[test]
+fn does_not_desugar_text() {
+    let spair = SPair("\\footnote{foo}".into(), S1);
+    let toks = vec![Nir::Text(spair)];
+
+    assert_eq!(
+        Ok(vec![Object(Nir::Text(spair))]),
+        Sut::parse(toks.into_iter()).collect(),
+    );
+}
+
 fn expect_expanded_header(
     sut: &mut Parser<InterpState, std::vec::IntoIter<Nir>>,
     given_val: &str,