fix: Incorrect return path identification with try/finally

[tannguyencse19]

Summary

Update to: Only let the finally block override the try path analysis when the finally block itself always terminates, for example it always returns or always raises.

If the finally block can fall through, analyze the try body, else, and handlers as usual.

Fixes #1091

Test Plan

cargo test test_return_try_finally_explicit_return

[yangdanny97]

Hmm, the linked issue was fixed by #1118, but I actually don't think that PR is correct. Let me discuss with the team

[github-actions]

Diff from mypy_primer, showing the effect of this PR on open source code:

scipy (https://github.com/scipy/scipy)
- ERROR scipy/optimize/_tstutils.py:135:24-25: Cannot set item in `dict[Unknown, None]` [unsupported-operation]
- ERROR scipy/stats/_axis_nan_policy.py:238:37-56: No matching overload found for function `numpy._typing._ufunc._UFunc_Nin2_Nout1.__call__` called with arguments: (type[float64] | int | Unknown, float) [no-matching-overload]

typeshed-stats (https://github.com/AlexWaygood/typeshed-stats)
- ERROR src/typeshed_stats/gather.py:999:28-1002:10: `Pattern[Unknown] | @6887 | None` is not assignable to upper bound `bytes | str` of type variable `AnyStr` [bad-specialization]
+ ERROR src/typeshed_stats/gather.py:999:28-1002:10: `Pattern[Unknown] | @7027 | None` is not assignable to upper bound `bytes | str` of type variable `AnyStr` [bad-specialization]

[yangdanny97]

After discussions w/ the team, I realize I misunderstood the previous PR. That is an adequate fix for the original issue, so this PR should not be necessary - the added test case passes without it