FIX: Standardize image-text converter font default to None

[romanlutz]

Description

PyRIT's two image-text converters had inconsistent font-name defaults. AddImageTextConverter already defaulted font_name to None (using Pillow's bundled default), but AddTextImageConverter defaulted to "helvetica.ttf", which does not exist on Linux. On Linux that caused ImageFont.truetype("helvetica.ttf", ...) to raise OSError and log a spurious Cannot open font resource: helvetica.ttf. Using Pillow built-in default font. warning before falling back. This warning surfaced in the 3_image_converters notebook during release validation, and the mismatched defaults made the two converters behave differently across platforms.

This change makes AddTextImageConverter mirror AddImageTextConverter:

• Signature default changed from font_name: str = "helvetica.ttf" to font_name: str | None = None.
• .ttf extension validation is now gated on font_name is not None.
• Added the _font_load_failed = font_name is None pattern so _load_font goes straight to ImageFont.load_default(size=...) when None (no truetype attempt, no warning), while keeping the existing OSError -> load_default fallback for custom fonts.
• Docstring updated to match AddImageTextConverter.

On the pinned Pillow (>=12.2.0), ImageFont.load_default(size=...) returns a real scalable TrueType font, so rendering was already fine; this cleans up the warning and makes both converters cross-platform consistent. No matplotlib/DejaVu font lookup was introduced. AddImageTextConverter was left untouched (already correct).

Tests and Documentation

• Updated test_add_text_image_converter_initialization to construct with the default None, assert _font_name is None and a real FreeTypeFont, and assert (via caplog) that no Cannot open font resource warning is emitted. Kept the invalid-.ttf ValueError test and the nonexistent-custom-font fallback-warning test. Both converter test files pass (31 tests), including test_add_image_text_converter_equal_to_add_text_image, which now confirms both converters produce byte-identical output for the None default on all platforms (previously only matched on Linux via fallback).
• Re-ran the doc/code/converters/3_image_converters notebook via JupyText (jupytext --to ipynb --execute). It executes end to end with 0 error outputs and 0 occurrences of Cannot open font resource/helvetica. The notebook source uses the plain converter constructors with no font workaround; the converter fix is what makes the defaults run cleanly.