Make the Executor Lost recovery handling configurable.

**Is your feature request related to a problem or challenge? Please describe what you are trying to do.**
A clear and concise description of what the problem is. Ex. I'm always frustrated when [...]
(This section helps Arrow developers understand the context and *why* for this feature, in addition to the *what*)

The current implementation is more a batch style, we should make the failure handling configurable. It may not be desired to roll back and recompute results. In an interactive query system for example, it may be desirable to just fail queries when an executor is lost rather than consume more cluster resources.

**Describe the solution you'd like**
A clear and concise description of what you want to happen.

**Describe alternatives you've considered**
A clear and concise description of any alternative solutions or features you've considered.

**Additional context**
Add any other context or screenshots about the feature request here.

Start by locating the Executor Lost recovery handling and the existing rollback/recompute path. Define the configuration entry point and failure behavior for interactive versus batch queries, then add coverage for each mode; done when callers can select fail-only or recovery behavior and tests verify both.

Written by the indexing model from the issue text.