improve: Span Adjustments

[monoxgas]

• Make some metric key handling stricter.
• Move attributes for objects to refs instead of events.
• Add run-level autolog specs and inheritence. Better handling for metric names.

---

Generated Summary:

• Updated the Dreadnode class to replace default values for log_inputs and log_output with Inherited, allowing for more flexible logging behavior.
• Introduced an autolog parameter in several methods, enabling automatic logging of task inputs, outputs, and metrics if not explicitly specified.
• Modified logging behavior in Task execution to utilize the autolog setting, which impacts how metrics are recorded (e.g., success rate and execution count).
• Added handling for the autolog parameter in RunSpan, affecting logging functions like log_metric and log_input to utilize a given prefix for logs.
• Enhanced the ObjectRef dataclass to include an attributes field, which provides additional context to logged objects.
• Introduced a new Inherited type for better clarity and type management when dealing with inherited default values for logging parameters.
• Altered several spans in the code to append attributes, improving the metadata rich logging aspect of the tracing functionality.

This summary was generated with ❤️ by rigging

Autogenerated:

• Refactored logging behavior in Dreadnode to allow for inherited logging configurations.
• Updated log_inputs and log_output parameters to support Inherited type, defaulting to INHERITED.
• Introduced autolog functionality to RunSpan, enabling automatic logging of task inputs/outputs when unspecified.
• Removed default boolean values for log_inputs and log_output across several functions, enhancing flexibility.
• Adjusted logging methods in the Task class to respect the new configuration and to pay attention to the autolog parameter.
• Modified log_metric and log_input methods in RunSpan and TaskSpan to support prefixed metric logging, adding contextual clarity.
• Added an attributes field to the ObjectRef dataclass for better attribute management on logged objects.
• Overall, these changes aim to enhance the logging capabilities, enabling more dynamic and context-driven logging behavior in the Dreadnode framework.