fix(abstractions): raise ValueError for out-of-range timedelta strings

[Max Azatian (HardMax71)]

Overview

parse_timedelta_string and parse_timedelta_from_iso_format let OverflowError through when a duration is past timedelta.max, like "50001140846:00021" from the issue or "P1000000000D". The callers only catch ValueError. So after #664 the OverflowError still escaped from three places: a typed timedelta property in JSON (it crashed deserialization of the whole response), the form serializer's additional data, and the JSON writer's string check. Both functions now raise ValueError from the OverflowError. These values are handled like any other invalid duration: get_timedelta_value returns None, additional data keeps the raw string, and the writer raises its own "Invalid timedelta string value found" error.

Related Issue

Fixes #480

Notes

The json and form packages get this through abstractions, so it reaches them with the next abstractions release. I didn't touch their microsoft-kiota-abstractions>=1.11.1 lower bound. The text serializer's get_timedelta_value doesn't catch anything, so there it now raises ValueError, as it already does for any other invalid string.

Testing Instructions

• cd packages/abstractions && pytest tests/test_date_utils.py -k out_of_range: out-of-range ISO durations (one of them parses to float infinity) and an out-of-range hh:mm:ss string raise ValueError.
• cd packages/serialization/json && pytest -k out_of_range: get_timedelta_value returns None and write_timedelta_value raises "Invalid timedelta string value found for property diff".
• cd packages/serialization/form && pytest -k out_of_range: get_object_value keeps the value as a string in additional data.
• All 11 new tests fail on main. Full suites pass on Python 3.10, 3.13 and 3.14: abstractions 150, json 200, form 55. yapf, mypy and pylint (10/10) are clean on all three. isort only flags the import order at the top of date_utils.py, which is the same on main.
• The repro from my comment on JSON deserialization does not handle OverflowError #480 against this branch: both out-of-range values now give the same results as "not-a-duration".

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[Max Azatian (HardMax71)]

try-catch-rethrow is tbh quite an antipattern. I'd say in general refactoring of this stuff is needed, but as a separate PR, unrelated to #480

ready for review btw