Add AI Resilience Capabilities

[MikeAlhayek]

This pull request introduces a new standalone package, CrestApps.Core.AI.Resilience, which provides builder-based resilience (retry) middleware for Microsoft.Extensions.AI clients. The framework now applies a default retry policy for framework-owned chat and utility clients, and host-created clients can opt in using the new package. The documentation and sample code have been updated to reflect these changes, and Azure OpenAI provider settings now expose and document matching retry policy options.

Key changes:

Resilience and Retry Middleware

• Added the new CrestApps.Core.AI.Resilience package, which provides UseDefaultResilience() and UseResilience(...) builder extensions for chat, embeddings, image, speech-to-text, and text-to-speech clients, enabling opt-in retry policies with exponential backoff and jitter. [1] [2] [3] [4]
• Updated DataExtractionService and PostSessionProcessingService to wrap resolved chat clients with the default resilience policy using the new builder extensions. [1] [2] [3] [4] [5] [6] [7]

Documentation and Guidance

• Added a dedicated "AI Resilience" documentation page describing usage, default retry schedules, extension points, and streaming behavior.
• Updated error-handling and orchestration docs to clarify that framework-owned chat and planning paths now use the default retry policy, while host-created clients can opt in via the builder extensions. [1] [2] [3] [4]

Azure OpenAI Provider Enhancements

• Exposed Azure SDK retry settings (MaxRetryAttempts, RateLimitRetryDelay, etc.) under CrestApps:AI:AzureClient with defaults matching the framework's resilience policy, and updated documentation to show these settings and their defaults. [1] [2] [3]

Release Notes

• Documented the new resilience package, its default retry policy, and the changes to client and Azure OpenAI retry handling in the v1.0.0 changelog.

These changes improve reliability for AI operations, provide clear opt-in mechanisms for host-created clients, and ensure consistent retry behavior across the framework and Azure OpenAI integrations.