Giving some normative structure to commands for "settings"

[jugglinmike]

Currently, the commands in the "settings" module ("getSettings", "getSupportedSettings", and "setSettings") are all defined with the following text:

The remote end steps given session and command parameters are implementation-defined.

This seems overly ambiguous to me. While we are deferring decisions on specific settings that can be controlled using ARIA-AT Automation, I think the proposal can and should define the following:

• the return value for successful invocations
• the error value for failed invocations
• the effect of implementation-defined behavior (e.g. "take implementation-defined steps to change the specified setting to the specified value")

Are there any reasons not to constrain implementations along these lines?

[jugglinmike]

Resolved via gh-54.