Fix AccessDeniedException when using AmazonBedrockModel in async evaluations

[Fizza-Mukhtar]

Bug Fix: AmazonBedrockModel fails in async evaluation context

This PR fixes an issue where AmazonBedrockModel raises an
AccessDeniedException when used inside DeepEval’s async evaluation
pipeline, despite valid AWS credentials and confirmed Bedrock access.

Problem

• AmazonBedrockModel worked with direct boto3 calls
• Failed only inside DeepEval async evaluations
• Error: AccessDeniedException during Converse calls
• Root cause: async session / credential handling was not aligned with
  aiobotocore best practices in concurrent contexts

Solution

• Ensure aiobotocore session handling is async-safe
• Avoid credential loss during concurrent async evaluations
• Keep behavior unchanged for sync execution paths

Testing

• Verified behavior with mocked aiobotocore async sessions
• Confirmed no regression for non-Bedrock models

This brings Bedrock behavior in line with other async-supported LLM providers in DeepEval.
Fixes #1916

[vercel]

@Fizza-Mukhtar is attempting to deploy a commit to the Confident AI Team on Vercel.

A member of the Team first needs to authorize it.

[greptile-apps]

PR author is not in the allowed authors list.