Implements Python startup and internal error metrics for HMD.

[dom96]

Just some scaffolding to help count Python internal error metrics. Tested upstream.

[ask-bonk]

I've now reviewed all the changed files thoroughly. Let me analyze the PR:

Changes summary:

1. observer.h: Adds setPythonWorkersInternalError() virtual method to RequestObserver
2. pyodide.h / pyodide.c++: Adds reportPythonWorkersInternalError JSG method to WorkerFatalReporter that calls the observer
3. util.ts: PythonWorkersInternalError constructor now calls FatalReporter.reportPythonWorkersInternalError() (with try/catch)
4. python-entrypoint-helper.ts: Adds TEST_ONLY_THROW_PYTHON_WORKERS_INTERNAL_ERROR helper gated behind experimental flag
5. Type declarations updated for the new method and experimental compat flag

Let me verify that this follows existing patterns correctly and check for any issues:

• The C++ pattern follows the exact same structure as reportFatal/setWorkerFatal — consistent.
• The try/catch in the constructor is appropriate since reporting metrics should not break error construction.
• The observer method has a no-op default — correct pattern for RequestObserver.
• The TEST_ONLY_ helper is properly gated behind experimental compat flag.
• No backward compatibility concerns — purely additive changes (new virtual with default, new JSG method).
• No security concerns — the new method just sets a metric flag, no capability leak.
• No tests are added with this PR, but the TEST_ONLY_THROW_PYTHON_WORKERS_INTERNAL_ERROR helper suggests tests will be added separately (likely in the internal/production test suite).

The code is clean, follows existing patterns, and is purely additive. There are no actionable issues.

LGTM

github run

[github-actions]

The generated output of @cloudflare/workers-types matches the snapshot in types/generated-snapshot 🎉

[codspeed-hq]

Merging this PR will improve performance by 39.14%

⚡ 1 improved benchmark
✅ 71 untouched benchmarks
⏩ 129 skipped benchmarks¹

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
⚡	simpleStringBody[Response]	27.2 µs	19.5 µs	+39.14%

---

_{Comparing dominik/python-hmd-metrics (789e44e) with main (764b4e8)}

Footnotes

1. 129 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩