feat(best-format): Support best format

[nakamuraos]

This pull request introduces a new "best format" selection feature for image output, allowing the system to automatically choose the most efficient image format based on content and configuration. It adds support for the format=best option, exposes several new environment variables for fine-tuning the behavior, and updates the configuration, parsing, and service layers accordingly. Documentation and tests have also been updated to reflect these changes.

Best format selection feature:

• Added support for format=best in image requests, enabling automatic multi-format encoding and selection of the smallest result. The list of candidate formats is configurable and lossless formats are included/excluded based on image complexity. [1] [2] [3]
• Introduced new environment variables (BEST_FORMAT_COMPLEXITY_THRESHOLD, BEST_FORMAT_MAX_RESOLUTION, BEST_FORMAT_BY_DEFAULT, BEST_FORMAT_ALLOW_SKIPS, BEST_FORMAT_PREFERRED_FORMATS) to control best format selection, with corresponding logic to parse and apply these settings. [1] [2] [3] [4] [5]

Configuration and service integration:

• Extended the Configuration struct and related logic to include a BestFormatConfig section, ensuring all services and tests construct and use the new configuration. [1] [2] [3] [4] [5] [6] [7] [8] [9] [10]

Proxy and pipeline logic:

• Updated the proxy service logic to trigger the best-format pipeline when format=best is requested or when the BEST_FORMAT_BY_DEFAULT setting is enabled and no format is specified. [1] [2]
• Added a method to TransformParams to distinguish between format-only transforms and those requiring re-encoding, supporting the BEST_FORMAT_ALLOW_SKIPS optimization.

Documentation and tests:

• Updated documentation (README.md, .env.sample) to describe the new feature, its configuration options, and usage. [1] [2] [3]
• Added and updated tests to cover configuration parsing, environment variable handling, and transform parameter logic for the new best format feature. [1] [2]

These changes collectively enable flexible, automatic selection of the optimal image format for each request, with robust configuration and clear documentation.