Robustness: Defensive NoneType handling during recordset or connection iteration

First, thank you so much for maintaining loguru. It's an indispensable tool in our stack, and I really appreciate the care put into its design.

Summary

While working with custom sinks and handling edge cases where a record or connection object might be uninitialized or return None, I noticed that certain serialization or iteration flows can trigger an unhandled AttributeError. Adding a light defensive check for NoneType would make custom sink integrations a bit more resilient and yield a clearer error message.

Environment

• Python version: 3.12.1
• Operating System: Ubuntu 22.04
• Loguru version: 0.7.2 (current main)

Steps to Reproduce

from loguru import logger

def awkward_sink(message):
    # Simulated edge case where record metadata might be empty/None
    record = message.record
    record["extra"] = None
    
    # Attempting to iterate or serialize when an inner attribute is None
    for _ in record["extra"]:
        pass

logger.add(awkward_sink)
logger.info("Triggering sink")

Expected Behavior

Loguru handles the edge case gracefully or raises a descriptive, context-aware error explaining that the target object is uninitialized or None.

Actual Behavior

An unhandled AttributeError: 'NoneType' object is not iterable bubbles up from deep within the handling/formatting sequence rather than a clean loguru-specific error.

Happy to submit a quick PR for this if the maintainers are open to it!