fix(observe): defer span end when a cancellation never resumed the generator

[DavidTraina]

What does this PR do?

Alternative to #1791 — same bug, see there for details and a repro. Maintainers should pick one; I'll close the other.

Includes the same core fix: aclose()/close() always close the wrapped generator, even if its span already ended, plus the same except TypeError narrowing.

On top of that: when a cancellation ends iteration before the generator ever resumed, don't end the span right away. Hold the error, close the generator first — so span updates in its finally still land — then end the span with the held error. The span outcome is the same ERROR/"CancelledError" as today; the difference is the generator's final updates are no longer dropped.

Scope: async generators only, CancelledError only, and only when the generator provably never resumed. Works on Python 3.10+ (inspect.getasyncgenstate is 3.12+, so it falls back to the ag_frame/ag_running attributes it reads).

Type of change

• Bug fix
• New feature
• Breaking change
• Refactor
• Documentation update
• Tooling, CI, or repo maintenance

Verification

uv run --frozen pytest -n auto --dist worksteal tests/unit   # also run on 3.10 and 3.11
uv run --frozen ruff check .
uv run --frozen ruff format --check .
uv run --frozen mypy langfuse --no-error-summary

Checklist

• I self-reviewed the diff using code_review.md.
• I added or updated tests for behavior changes.
• I updated docs, examples, or .env.template if needed.
• I did not hand-edit generated files; if generated files changed, I used the upstream regeneration path.
• I did not commit secrets or credentials.

[claude]

Claude Code Review

This pull request is from a fork — automated review is disabled. A repository maintainer can comment @claude review to run a one-time review.

[hassiebp]

Thanks for your contribution, @DavidTraina !