Aware datetime offsets are lost in SLO history timestamp conversion #1013

Describe the bug

force_to_epoch_seconds() ignores aware datetime offsets. This affects the
time-range parameters built by ServiceLevelObjective.history() and conversion
of datetime inputs through parse_date_as_epoch_timestamp().

To Reproduce

from datetime import datetime, timedelta, timezone
from datadog.util.format import force_to_epoch_seconds

utc = datetime(2024, 1, 1, tzinfo=timezone.utc)
india = datetime(2024, 1, 1, 5, 30, tzinfo=timezone(timedelta(hours=5, minutes=30)))
print(force_to_epoch_seconds(utc))    # 1704067200
print(force_to_epoch_seconds(india))  # 1704087000

Expected behavior

These represent the same instant; both should return 1704067200. Currently the
helper passes the local wall-clock tuple to calendar.timegm, losing the offset.

Environment and Versions

datadog 0.55.0 from main a483dd9; Windows / Python 3.12.14.

Additional context

A local change using utctimetuple preserves existing naive-UTC, numeric and
whole-second behavior. Ten new failing cases become passing; 96 focused
utility/mocked API tests pass. Broader import-isolation failures and native
Windows typing findings are also reproduced upstream. No live Datadog service
was contacted. A focused patch and tests are prepared if this approach is welcome.

Prepared and locally validated with OpenAI Codex assistance. No production
incident or independent human review claimed. Severity/triage left to maintainers.