Period constructed from invalid high-resolution Timestamp should raise OutOfBoundsDatetime

[pavansai018]

Summary

Constructing a Period from a Timestamp whose underlying integer becomes out-of-range when interpreted at a higher-resolution frequency (e.g., treating Timestamp.min.value as microseconds) previously triggered a low-level OverflowError inside get_period_ordinal, which is declared noexcept. This resulted in an unraisable exception and Period(...) silently returning an incorrect ordinal instead of raising OutOfBoundsDatetime.

What this PR changes

A new internal helper _period_ordinal_safe(...) is introduced. It mirrors the logic of get_period_ordinal but is allowed to raise Python exceptions. cpdef period_ordinal(...), which feeds the Python-facing Period(...) API, now delegates to this safe helper.

This ensures:

• low-level period arithmetic and internal callers continue using the original get_period_ordinal unchanged,
• Python-facing constructions such as Period(Timestamp, "us") correctly raise OutOfBoundsDatetime when the timestamp cannot be represented at the target frequency,
• no C-level OverflowError leaks into pytest as an unraisable exception.

Test

The regression test:

ts = pd.Timestamp(pd.Timestamp.min.value, unit="us")
with pytest.raises(OutOfBoundsDatetime):
    pd.Period(ts, freq="us")

Fixes #63278

[jbrockmendel]

We don't want this to raise.