feat(usage): track cached tokens + correct input/output/cache cost (#2209)

Normalize every provider to one cache-inclusive convention via
canonicalizeUsage() before persist, and price cached + cache_creation as
subsets of prompt_tokens in calculateCostFromTokens() to stop
double-counting. usageRepo now delegates cost math to a single source.
Surface Cached tokens/cost across dashboard (overview, tokens, cost,
details). Merge Claude message_start cache with message_delta output so
cache counts survive. Compatible LLM nodes now allow multiple API-key
connections (key pool).

Co-authored-by: Cursor <cursoragent@cursor.com>

open-sse/translator/concerns/usage.js

@@ -39,7 +39,16 @@ const USAGE_EXTRACTORS = {
   },
   kiro(raw) {
     const input = n(raw.inputTokens), output = n(raw.outputTokens);
-    return { promptTokens: input, completionTokens: output, totalTokens: input + output };
+    // ponytail: Amazon Q (Kiro upstream) does not expose cache fields today,
+    // but pass through any cache_read/cache_creation/cached_tokens if the
+    // event shape grows them later so cost tracking keeps working without
+    // a second pass.
+    const cached = n(raw.cache_read_input_tokens) || n(raw.cachedTokens) || n(raw.cached_tokens);
+    const cacheCreation = n(raw.cache_creation_input_tokens);
+    const out = { promptTokens: input, completionTokens: output, totalTokens: input + output };
+    if (cached > 0) out.cachedTokens = cached;
+    if (cacheCreation > 0) out.cacheCreationTokens = cacheCreation;
+    return out;
   },
   ollama(raw) {
     const input = n(raw.prompt_eval_count), output = n(raw.eval_count);