improve openai instrumentation

[tpatel]

• capture all settings
• only support openai>=1.0.0
• add a check_all_requirements util to easily check if the instrumentation requirements are available
• simplify the monkey-patching logic to enable instrumentation of other modules
• add testing to validate the monkey-patching
• fixed some typing errors
• add streaming support

There is one remaining feature that need implementation:

• image input (the message content property isn't text in that case, it's an array instead, we need to handle that)