Improve first-run experience and event store selection

[woksin]

Added

• Running cratis with no arguments now shows the active context name and server URL, so it is always clear where the CLI is pointed.
• Connection error messages now include the active context name, server URL, and the exact commands to update the connection (cratis context set-value server / cratis context create).
• cratis llm-context --schema outputs the JSON Schema for the llm-context format — useful for tooling that validates or generates against the descriptor shape.

Changed

• First run automatically creates a default context pointing at chronicle://localhost:35000/?disableTls=true — no manual cratis context create step required.
• The event store prompt now also fires when the stored event store no longer exists on the server (e.g. it was deleted), re-prompting the user and saving the new selection before the command runs.
• cratis llm-context now outputs a proper nested hierarchy: chronicle sub-groups (observers, jobs, events, etc.) appear as subGroups instead of being flattened alongside top-level commands.
• cratis llm-context hoists -e -n to the chronicle group level — they appear once in inheritedOptions on the group rather than being repeated on every sub-group and command.
• cratis llm-context now separates positional arguments (e.g. <OBSERVER_ID>) from named options (e.g. --type) in the command descriptor, making it unambiguous how each parameter is supplied.
• All commands now carry AI-focused descriptions in cratis llm-context output, distinct from their --help text, with more context on when and how to use them.

Fixed

• cratis llm-context no longer includes examples in the output.
• Angle brackets in cratis llm-context JSON output are now emitted as literal </> instead of \u003C/\u003E.
• PowerShell completion script now correctly quotes 'cratis' in Register-ArgumentCompleter -CommandName.