Clean, central record of all job step scripts' commands

The proposed feature is a logging mechanism for what commands are actually issued to the system.

In its simplest form, this could be imagined as a single bash script that shows what steps are actually executed (or would be executed, if we imagine that this could be obtained before the study is launched).

This would be valuable for two reasons:

1. It records what was actually executed; this contrasts with the yaml specification, which programmatically describes what is intended to be executed.
2. It provides a means for a user to understand the translation of a Maestro specification to the actual execution; everything unpacked and put in a centralized location.

Frank's suggestions as to where this might be implemented are:

1. In start_process
2. Logging into the metadata via each adapter.

Either of these would provide the opportunity to catch and record the commands.

An alternate idea, as mentioned above, would be to composite the whole specification and its scattered job scripts into one large script before launching the study; say, via a special "dry-run" adapter, which would pass through the study structure, pulling out all of the scripts and effectively pasting them together.

Start by reading start_process and the adapter implementations mentioned in the issue. Compare recording commands there with logging them into metadata, and consider how a dry-run adapter would traverse the study structure. Done means the commands actually issued, or the commands that would be issued, are captured in one central record.