feat: Enhance AsyncPlaywrightCrawlerStrategy with text-only and light modes, dynamic viewport adjustment, and session management

### New Features:
- **Text-Only Mode**: Added support for text-only crawling by disabling images, JavaScript, GPU, and other non-essential features.
- **Light Mode**: Optimized browser settings to reduce resource usage and improve efficiency during crawling.
- **Dynamic Viewport Adjustment**: Automatically adjusts viewport dimensions based on content size, ensuring accurate rendering and scaling.
- **Full Page Scanning**: Introduced a feature to scroll and capture dynamic content for pages with infinite scroll or lazy-loading elements.
- **Session Management**: Added `create_session` method for creating and managing browser sessions with unique IDs.

### Improvements:
- Unified viewport handling across contexts by dynamically setting dimensions using `self.viewport_width` and `self.viewport_height`.
- Enhanced logging and error handling for viewport adjustments, page scanning, and content evaluation.
- Reduced resource usage with additional browser flags for both `light_mode` and `text_only` configurations.
- Improved handling of cookies, headers, and proxies in session creation.

### Refactoring:
- Removed hardcoded viewport dimensions and replaced them with dynamic configurations.
- Cleaned up unused and commented-out code for better readability and maintainability.
- Introduced defaults for frequently used parameters like `delay_before_return_html`.

### Fixes:
- Resolved potential inconsistencies in viewport handling.
- Improved robustness of content loading and dynamic adjustments to avoid failures and timeouts.

### Docs Update:
- Updated schema usage in `quickstart_async.py` example:
  - Changed `OpenAIModelFee.schema()` to `OpenAIModelFee.model_json_schema()` for compatibility.
- Enhanced LLM extraction instruction documentation.

This commit introduces significant enhancements to improve efficiency, flexibility, and reliability of the crawler strategy.

docs/examples/quickstart_async.py

@@ -128,7 +128,7 @@ async def extract_structured_data_using_llm(provider: str, api_token: str = None
             extraction_strategy=LLMExtractionStrategy(
                 provider=provider,
                 api_token=api_token,
-                schema=OpenAIModelFee.schema(),
+                schema=OpenAIModelFee.model_json_schema(),
                 extraction_type="schema",
                 instruction="""From the crawled content, extract all mentioned model names along with their fees for input and output tokens. 
                 Do not miss any models in the entire content. One extracted model JSON format should look like this: