This PR fixes how add_vline() handles annotations, addressing the bug described in #3065
where add_vline(x=, annotation_text=...) caused a TypeError.
It refactors the logic to use the new shape.label attribute instead of creating a separate annotation, which removes the arithmetic step that caused the failure.

Why this change
Previously, add_vline created both a shape and an annotation to display text.
To position the annotation, it computed the mean/min/max of the line’s coordinates — which broke for mixed data types (e.g., strings + datetimes).
Using shape.label removes that calculation entirely and simplifies the API.

What’s included

Added helper _coerce_shape_label_from_legacy_annotation_kwargs(kwargs)

• Safely maps legacy annotation_* kwargs (text, font, textangle) into a shape.label dict, emitting a FutureWarning.

Added _normalize_legacy_line_position_to_textposition()

• Maps legacy annotation_position to "start" | "middle" | "end".

Updated _process_multiple_axis_spanning_shapes()

• For vline, creates a single labeled shape (label.textposition maps: top→end, bottom→start, middle→middle).
• Skips creating a layout.annotation

Feedback

• @DatBoiWhoWon @lilyMaung @chanlii please check the functions and helpers to see if we agree on this approach for the other methods.