Add instrumentation for `Images.generate()` and `Images.edit()`

The OpenAI integration does not instrument image generation endpoints. These are commonly used for DALL-E and generate meaningful cost/performance telemetry.

**Endpoints to instrument:**

* `openai.resources.Images.generate()` (sync + async)
* `openai.resources.Images.edit()` (sync + async)

**Attributes to capture:**

* Model name
* Prompt / input text (if `send_default_pii` and `include_prompts` are enabled)
* Image size, quality, and style parameters
* Number of images requested (`n`)
* Response metadata (created timestamp, revised prompt if returned)
* Follow OTel GenAI semantic conventions for span naming and attributes

**Reference:** [OpenAI integration compliance review](<../openai-integration-compliance-review.md>), Section 6