[Agent Builder] Skip esql specs in agent-builder evals run (elastic#267522)

## Summary

- Adds `testIgnore: ['**/esql/**']` to the agent-builder Playwright eval
config so the recursive `testDir` walk no longer picks up
`evals/esql/esql.spec.ts`.
- That spec is owned by the separate `esql-generation` suite
(`esql.playwright.config.ts`), so today it runs twice every weekly LLM
evals cycle — ~4 min/run × ~7 connectors = **~28 min/week of duplicate
work** in the agent-builder child.
- Same recursive-`testDir` bug pattern fixed for entity-analytics v1 in
elastic#267451.

Refs elastic/security-team#17139

## Stacking

Stacked on top of elastic#267451 (`evals/ea-v1-ignore-v2-specs`), which extends
`createPlaywrightEvalsConfig` to forward the optional `testIgnore`. Will
retarget to `main` once elastic#267451 merges.

## Test plan

- [ ] `node scripts/type_check --project
x-pack/platform/packages/shared/agent-builder/kbn-evals-suite-agent-builder/tsconfig.json`
clean
- [ ] `node scripts/eslint
x-pack/platform/packages/shared/agent-builder/kbn-evals-suite-agent-builder/playwright.config.ts`
clean
- [ ] Next weekly `kibana-evals-weekly-llm-evals` run: agent-builder
Playwright report no longer lists `evals/esql/esql.spec.ts`, p50 drops
by ~4 min vs `build_13` baseline
- [ ] `esql-generation` suite continues to pass on its own config

🤖 Generated with [Claude Code](https://claude.com/claude-code)

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Co-authored-by: Elastic Machine <elasticmachine@users.noreply.github.com>

Author: 3 people

x-pack/platform/packages/shared/agent-builder/kbn-evals-suite-agent-builder/playwright.config.ts

@@ -9,6 +9,11 @@ import { createPlaywrightEvalsConfig } from '@kbn/evals';
 
 export default createPlaywrightEvalsConfig({
   testDir: Path.resolve(__dirname, './evals'),
+  // The `esql-generation` suite (`esql.playwright.config.ts`) owns specs
+  // under `evals/esql/`. Excluding them here prevents the recursive
+  // `testDir` walk from picking them up and double-running them in the
+  // agent-builder weekly cycle.
+  testIgnore: ['**/esql/**'],
   // CI job timeout is ~1h; keep default low and use EVALUATION_REPETITIONS
   // for longer/higher-confidence runs.
   repetitions: 1,

x-pack/platform/packages/shared/kbn-evals/src/config/create_playwright_eval_config.ts

@@ -24,11 +24,13 @@ export interface EvaluationTestOptions extends ScoutTestOptions {
  */
 export function createPlaywrightEvalsConfig({
   testDir,
+  testIgnore,
   repetitions,
   timeout,
   runGlobalSetup,
 }: {
   testDir: string;
+  testIgnore?: PlaywrightTestConfig['testIgnore'];
   repetitions?: number;
   timeout?: number;
   runGlobalSetup?: boolean;
@@ -106,5 +108,6 @@ export function createPlaywrightEvalsConfig({
     globalSetup: require.resolve('./setup.js'),
     globalTeardown: require.resolve('./teardown.js'),
     timeout: timeout ?? 5 * 60_000,
+    ...(testIgnore !== undefined ? { testIgnore } : {}),
   });
 }