Fix add_vline and add_hline with datetime axes

[mosh3eb]

Summary

When using add_vline, add_hline, add_vrect, or add_hrect with annotation_text on datetime axes, plotly crashes with a TypeError because the _mean() helper function tries to call float(sum(x)) on datetime strings.

Approach

Instead of assuming input values are equal (which only works for vlines/hlines), _mean() now converts datetime-like values to milliseconds-since-epoch, computes the actual arithmetic mean, and converts back. This correctly handles annotation placement for ALL shapes with datetime axes, including rects where x0 != x1.

• The numeric fast path is unchanged -- datetime handling only activates when sum() raises TypeError
• Handles both date strings ("2018-09-24") and datetime.datetime objects
• Re-raises TypeError for truly unsupported types

Changes

• plotly/shapeannotation.py: Added _is_date_string(), _datetime_str_to_ms(), _ms_to_datetime_str() helpers and updated _mean() with try/except fallback
• tests/test_optional/test_autoshapes/test_annotated_shapes.py: Added 5 tests covering datetime strings, datetime objects, vlines, hlines, and vrects

Fixes #3065

[emilykl]

Thanks for the contribution @mosh3eb ! This approach won't work because the _mean() function is used to determine annotation placement for ALL lines, not just vertical/horizontal lines, so we can't assume that the input values are equal. See the linked issue for further discussion of possible fixes.