A Lambda function processes objects from an S3 bucket but some invocations are slower than others due to initialization (loading libraries, creating clients). The developer needs predictable, low-latency invocation durations and requires that initialization work occur during allocation time rather than during each invocation. Which combination of steps will achieve this? (Choose two.)

Enabling provisioned concurrency for the Lambda function ensures that a specified number of execution environments are pre-initialized and ready to respond to invocations. This eliminates cold starts for those pre-warmed instances, leading to predictable, low-latency invocation durations as initialization work (loading libraries, creating clients) occurs during the allocation of these environments, not during each invocation. Deploying changes and publishing a new, immutable version of the Lambda function is a standard best practice for managing Lambda code. It allows for controlled deployments, easy rollbacks, and ensures that the provisioned concurrency is applied to a specific, stable version of the function. Using the $LATEST version is generally not recommended for production as it's mutable and can change unexpectedly. Setting reserved concurrency limits the maximum concurrent invocations but doesn't pre-initialize environments. Creating a schedule group in Amazon EventBridge Scheduler is for scheduled invocations, not for optimizing cold starts.