[Fix]: Incompatibilities with OpenTelemetry LLM semantics pending release

[patcher9]

Overview:

This PR introduces changes that makes sure OpenLIT is following the OTel Semantic conventions for GenAI

resolves #342

Changes:

Changes in the SDK

• Trace attribute gen_ai.request.top_p is made a double
• Setting default value for Trace attribute gen_ai.request.max_tokens as -1 instead of a string
• Trace attribute gen_ai.request.temperature is made a double
• Trace attribute gen_ai.request.presence_penalty is made a double
• Trace attribute gen_ai.request.frequency_penalty is made a double
• Trace attribute gen_ai.response.finish_reason' is renamed to 'gen_ai.response.finish_reasons'
• Trace attribute gen_ai.usage.completion_tokens is renamed to gen_ai.usage.output_tokens
• Trace attribute gen_ai.usage.prompt_tokens is renamed to gen_ai.usage.input_tokens

Changes in the UI

• Trace attributes gen_ai.response.finish_reason' is renamed to 'gen_ai.response.finish_reasons'
• Trace attributes gen_ai.usage.completion_tokens is renamed to gen_ai.usage.output_tokens
• Trace attributes gen_ai.usage.prompt_tokens is renamed to gen_ai.usage.input_tokens

Visuals (If applicable):

NA

Checklist:

• PR name follows conventional commit format: [Feat]: ... or [Fix]: ....
• Added visuals for changes (If applicable)
• Checked OpenLIT contribution guidelines