perf: the word lists travel as .txt, which the host compresses

GitHub Pages serves a .dic as text/x-c and leaves it uncompressed, while
text/plain goes out gzipped: the English list cost 552 KB over the wire
where 191 KB would do. Only the fetch path changes; the mount name the
checker sees stays <code>.dic. Measured first load 1.83 → 1.70 MB.

src/lib/spell/dictionary.ts

@@ -29,13 +29,14 @@ async function fetchBytes(url: string): Promise<Uint8Array> {
 }
 
 async function build(code: string): Promise<Checker> {
-  // Vendored assets: public/dictionaries/<code>/<code>.{aff,dic}. BASE_URL keeps
-  // this correct under a non-root deploy base.
+  // Vendored assets: public/dictionaries/<code>/<code>.{aff,dic.txt}. The word list is
+  // .txt so the host gzips it (GitHub Pages leaves a .dic's text/x-c uncompressed);
+  // BASE_URL keeps the path correct under a non-root deploy base.
   const base = `${import.meta.env.BASE_URL}dictionaries/${code}/${code}`;
   const [factory, aff, dic] = await Promise.all([
     loadFactory(),
     fetchBytes(`${base}.aff`),
-    fetchBytes(`${base}.dic`),
+    fetchBytes(`${base}.dic.txt`),
   ]);
   const affPath = factory.mountBuffer(aff, `${code}.aff`);
   const dicPath = factory.mountBuffer(dic, `${code}.dic`);

tests/unit/spell.test.ts

@@ -14,7 +14,7 @@ function dictPath(rel: string): string {
 async function makeChecker(code: string): Promise<Hunspell> {
   const factory = await loadModule();
   const aff = factory.mountBuffer(readFileSync(dictPath(`${code}/${code}.aff`)), `${code}.aff`);
-  const dic = factory.mountBuffer(readFileSync(dictPath(`${code}/${code}.dic`)), `${code}.dic`);
+  const dic = factory.mountBuffer(readFileSync(dictPath(`${code}/${code}.dic.txt`)), `${code}.dic`);
   return factory.create(aff, dic);
 }