Optimize reference resolution with in-memory caches

The resolving refs phase stalled on large projects (3400+ files, 38k+ nodes)
because matchFuzzy loaded ALL functions/methods/classes per ref, import
mappings were re-extracted per ref, and fileExists hit disk every call.

Add kindCache, lowerNameCache, importMappingCache, and knownFiles set to
warmCaches(). Rewrite matchFuzzy to use O(1) lowercase index lookup instead
of 3x getNodesByKind scans. Cache import mappings per file. Pre-build file
existence set from the index for O(1) fileExists checks.

src/resolution/import-resolver.ts

@@ -432,14 +432,12 @@ export function resolveViaImport(
   ref: UnresolvedRef,
   context: ResolutionContext
 ): ResolvedRef | null {
-  // Read the source file to extract imports
-  const content = context.readFile(ref.filePath);
-  if (!content) {
+  // Use cached import mappings (avoids re-reading and re-parsing per ref)
+  const imports = context.getImportMappings(ref.filePath, ref.language);
+  if (imports.length === 0 && !context.readFile(ref.filePath)) {
     return null;
   }
 
-  const imports = extractImportMappings(ref.filePath, content, ref.language);
-
   // Check if the reference name matches any import
   for (const imp of imports) {
     if (imp.localName === ref.referenceName || ref.referenceName.startsWith(imp.localName + '.')) {