Instrument `generate_images` / `edit_image` / `upscale_image`

## Problem

Image generation is a significant API surface on the google-genai library that is not instrumented:

* `Models.generate_images` — text-to-image (e.g. Imagen 3)
* `Models.edit_image` — edit with reference images (mask, style, subject, etc.)
* `Models.upscale_image` — upscale x2 or x4
* `Models.recontext_image` — product recontext / virtual try-on (Vertex AI only)
* `Models.segment_image` — image segmentation/masking (Vertex AI only)

No `gen_ai.generate_image` operation exists in OTel semconv yet, so a span operation will need to be defined.

## Files

* `sentry_sdk/integrations/google_genai/__init__.py` — `setup_once()` needs additional patches
* `sentry_sdk/consts.py` — may need new `OP` constant(s)

## Acceptance Criteria

* Image generation/editing calls produce spans with appropriate operation names
* Async variants are also instrumented
* Model name and prompt are captured (prompt respects PII settings)
* Tests verify span creation

Start in sentry_sdk/integrations/google_genai/__init__.py by reading setup_once() and its existing patches, then review sentry_sdk/consts.py for operation constants. Trace the synchronous and asynchronous image methods named in the issue, and verify that calls create spans with operation names, model and PII-aware prompt data; add tests that cover span creation.

Escrita pelo modelo de indexação a partir do texto da issue.