Full boto-like credential configuration support

Trying out a Nova Speech-to-Speech example today, I was surprised `aws_sdk_bedrock_runtime` didn't correctly pick up my configured AWS credentials - with the below error:

```
aws_sdk_bedrock_runtime.models.ServiceError: AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY are required
```

[This note in the readme](https://github.com/awslabs/aws-sdk-python/blob/0090493b47493aea886c10ee9626755b36481f12/dev-guide/service-clients.md?plain=1#L21) seems to suggest there's a known limitation here - but is there an expected timeline for this library to align to full standard CLI/boto3 credential discovery?

It was straightforward enough for me to work around on my corp laptop, but I'm worried that there are AWS environments e.g. SageMaker and [ECS](https://docs.aws.amazon.com/AmazonECS/latest/developerguide/ecs-environment-variables.html) where AWS credentials are available and people might want to run this connector, but AFAIK these specific environment variables are not set?

Start with the linked service-clients README note and reproduce the credential error in the Nova Speech-to-Speech example. Check how aws_sdk_bedrock_runtime currently obtains credentials, then compare the expected standard CLI/boto3 discovery behavior for SageMaker and ECS. Done means supported configured AWS credentials are discovered without requiring AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY.

索引モデルが issue の本文から書いたものです。