test: add regression test for issue #234 prompt caching behavior

[robert-j-y]

Description

Adds a regression test for issue #234 ("Delayed caching with GPT 4.1, none with Kimi K2 Thinking") to the e2e/issues/ directory.

What the test verifies:

• Provider correctly exposes cachedTokens in providerMetadata.openrouter.usage.promptTokensDetails when present
• Provider metadata includes detailed usage information (promptTokens, completionTokens, totalTokens)
• Response structure is valid regardless of whether caching occurs
• For Kimi K2 (thinking model), also verifies reasoningTokens is correctly exposed

Test structure:

• Makes 3 consecutive requests for both GPT 4.1 and Kimi K2 (matching the exact issue scenario)
• Logs output in a concise single-line format per request
• Test header documents only the known facts from the issue (no root cause speculation)

Updates since last revision

• Removed redundant comments throughout the test file
• Consolidated console.log output to single-line format per request
• Test code is now minimal and focused on verification logic

Human Review Checklist

• Verify test follows the e2e/issues/ pattern from PR test: add e2e/issues directory with issue #194 regression test #369
• Confirm test header documents only known facts from the issue
• Check if 180s timeout is reasonable for 6 API calls (3 GPT 4.1 + 3 Kimi K2)
• Note: Tests verify structure, not caching behavior (caching is non-deterministic and depends on server routing)
• Note: Tests depend on external APIs - may be flaky if models are unavailable

Checklist

• I have run pnpm stylecheck and pnpm typecheck
• I have run pnpm test and all tests pass
• I have added tests for my changes (if applicable)
• I have updated documentation (if applicable) - N/A

Changeset

• I have run pnpm changeset to create a changeset file

Note: Empty changeset used since this is a test-only change that doesn't need a release.

---

Link to Devin run: https://app.devin.ai/sessions/7ed009fd989445428e2bd9f81b697781
Requested by: Robert Yeakel (@robert-j-y)