bench: drop flaky isinstance, event handler, and compiler import benchmarks

[FarhanAliRaza]

What

Removes CodSpeed benchmarks that flag noise as regressions on unrelated PRs.

Removed

• test_isinstance_container[list_dict] (tests/benchmarks/test_isinstance.py): validated a 10k-element list[dict[str, int]] through _isinstance(nested=1). It exercised the same per-element recursion as the remaining list_int, list_optional and list_typeddict params, so no coverage is lost.
• test_from_event_type[event_handler] (tests/benchmarks/test_event_creation.py): called Event.from_event_type with a bare EventHandler. That path calls the handler to build an EventSpec, which is already covered by the event_spec and lambda_event_handler params.
• test_from_event_type[lambda_event] (same file): called Event.from_event_type with a lambda returning an existing Event. It drifted 3-4% on unrelated PRs (ENG-12486 feat(sdk): log in, review code, and configure cloud providers #7175 and this one). The event param covers the Event pass-through branch and lambda_event_spec / lambda_event_handler cover the lambda unwrapping branch.
• test_from_event_type[lambda_event_spec] (same file): called Event.from_event_type with a lambda returning an EventSpec. It drifted 3-4% on ENG-12486 feat(sdk): log in, review code, and configure cloud providers #7175 and on this PR. It is the event_spec path plus one lambda call, which lambda_event_handler already covers.
• test_get_all_imports and test_collect_imports (tests/benchmarks/test_compilation.py): measured import collection alone, once via the legacy Component._get_all_imports recursion and once via a plugin walk restricted to imports. Both _complicated_page variants drift by several percent between runs. test_compile_page and test_compile_all_artifacts still walk the tree and collect imports as part of the full compile, and test_import_reflex keeps measuring cold import reflex time.

The import-only collector plugin in tests/benchmarks/fixtures.py had no other user and is removed with them.

Testing

uv run pytest tests/benchmarks
46 passed

No package source is touched, so this carries skip-changelog.

[cubic-dev-ai]

No issues found across 4 files

_{Re-trigger cubic}

[greptile-apps]

Confidence Score: 5/5

The benchmark-only changes appear safe to merge, with one non-blocking stale-docstring issue remaining in the existing review thread.

Findings

1.  Stale Coverage Description ▶

Summary

This PR removes noisy CodSpeed benchmark cases and the import-only fixture that became unused.

• Removes redundant _isinstance coverage for a large list[dict].
• Removes flaky event-normalization benchmark parameters.
• Removes standalone import-collection benchmarks while retaining full compilation benchmarks.
• Removes the now-unused import collector plugin and associated imports.

_{Reviews (4) · Last reviewed commit: "bench: drop the flaky lambda_event_spec ..."}

[codspeed]

Merging this PR will not alter performance

✅ 45 untouched benchmarks
⏩ 17 skipped benchmarks¹

---

_{Comparing FarhanAliRaza:farhan/remove-flaky-benchmarks (02ab2c5) with main (00228b8)}

Footnotes

1. 17 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. ↩