Add unit tests for BedrockClassifier in Python

[amandladev]

The BedrockClassifier implementation currently has no dedicated test coverage. The existing classifier tests only cover the generic Classifier base class and the AnthropicClassifier.
Since BedrockClassifier is the AWS-native classifier and a core component of the framework, it should have parity with the Anthropic test suite.

Proposed Changes
Add test_bedrock_classifier.py covering:

• Initialization — client injection, auto-creation via boto3, AWS_REGION env fallback, custom model ID, inference coding defaults/overrides, user_agent registration.
• Tool Schema — Bedrock toolSpec structure validation.
• toolChoice branching — correctly added for Anthropic and Mistral Large model IDs, omitted for others.
• Happy path — successful classification, agent selection, confidence parsing.
• Error handling — no output, missing/empty toolUse, invalid tool input, BotoCoreError, ClientError.
• Callbacks — on_classifier_start / on_classifier_stop invocation and argument verification.
• Edge cases — non-existing agent, string-to-float confidence coercion, multiple content blocks, system prompt list format.