Restrict EC2CreateInstanceOperator cleanup to waiter failures and add guard flag

[SameerMesiah97]

Description

This PR follows up on PR #60904 by refining cleanup behavior in EC2CreateInstanceOperator. Cleanup behavior is guarded by a flag and is opted in by default. Cleanup is only triggered for post-start EC2 instance failures (including WaiterError), ensuring termination is attempted only when an instance was successfully created and avoiding interception of non-AWS exceptions.

Rationale

Restricting cleanup to post-creation EC2 instance failures prevents unintended termination in unrelated failure paths while still addressing orphaned instances created during execution, and aligns EC2 cleanup semantics with recent changes to other AWS resource creation operators proposed in PRs #61145, #61051, and #61010 .

Tests

Existing tests continue to validate cleanup behavior. The new terminate_instance_on_failure flag is now explicitly set to True in applicable tests for explicitness.

Documentation

The docstring for EC2CreateInstanceOperator has been updated with a brief description of the new flag term.

Backwards Compatibility

A new flag, terminate_instance_on_failure, has been added to EC2CreateInstanceOperator with a default value of True. Best-effort cleanup will now be attempted if a post-creation failure (including WaiterError) occurs after the EC2 instance has been successfully created.

[SameerMesiah97]

@vincbeck

Sorry for tagging you here but this is effectively the same pattern for the other resource-leaking AWS operators applied retrospectively to EC2CreateInstanceOperator. Since you reviewed them, I figure you might be interested in reviewing this PR.