testing mode

Right now I don't think the the design is well-suited for tests.

For global init in tests it's typical to have a global lazy_static that all tests try to use, ensuring global setup happens once. However we currently make setup and running a single transactional operation. Ideally we could have an alternate way of starting things that lets each test share global init that can be shared *AND* let each have their own "sandbox". Like it would be great if you could ask for logs that were generated by test1 without test2 messing with it.

idk how much of a pipe dream the full sandboxing stuff is, but at very least we want to be able to get the primary output/error of each test individually.

The issue names no files, tests, or entry points, so begin by locating the setup and transaction lifecycle in the Rust CLI library and reading its existing tests. Clarify whether the goal includes full sandboxing or only per-test primary output and error capture; done criteria are not defined beyond isolating those results.

Written by the indexing model from the issue text.