repl: update .scenario tab completion for group-qualified multi-API usage

[Copilot]

Summary

Refactors .scenario REPL completion to match group-qualified multi-API syntax while preserving single-API behavior. Multi-runner completion now resolves in two stages: group selection first, then group-scoped scenario path/function completion.

• Completion flow
  • .scenario <Tab> (multi-API): suggests configured API groups
  • .scenario <group> <Tab>: suggests scenario functions/path prefixes from that group’s scenario registry
  • Unknown groups return empty completions without throwing
• Compatibility
  • Existing single-runner .scenario <path> completion logic is retained
• Coverage
  • REPL completion tests now cover multi-runner group suggestions, group-scoped nested path completion, unknown-group fallback, and existing single-runner behavior
• Docs + release note
  • REPL docs updated to describe single vs multi-API tab completion behavior
  • Added patch changeset entry

// Multi-API completer wiring
createCompleter(
  rootBinding.registry,
  builtinCompleter,
  rootBinding.openApiDocument,
  rootBinding.scenarioRegistry,
  {
    billing: billingScenarioRegistry,
    inventory: inventoryScenarioRegistry,
  },
);

// Expected UX:
// .scenario <Tab>            -> ["billing", "inventory"]
// .scenario billing <Tab>    -> ["setup", "pets/", ...]

Original Prompt

This section details on the original issue you should resolve

Issue title: repl: update tab completion for group-qualified .scenario

Issue description:
Update REPL tab completion so .scenario completions reflect the new multi-API qualifier syntax while preserving the current single-API experience.

Context

Current completion assumes .scenario <path>. With multiple runners, users need completions for:

1. available group keys after .scenario
2. scenario paths/functions after .scenario <group>

Proposed change

Refactor the .scenario completer to operate in two stages for multi-runner mode and keep existing path/function completion for single-runner mode.

Acceptance criteria

• In multi-runner mode, typing .scenario offers available group keys
• In multi-runner mode, typing .scenario <group> offers scenario path/function completions scoped to that group
• Invalid groups do not crash completion and fall back gracefully
• In single-runner mode, existing .scenario <path> completions are preserved
• REPL tests assert completion behavior for both modes and nested scenario paths

Manual acceptance tests

• In a single-API REPL session, typing .scenario then pressing Tab suggests scenario functions and path prefixes from the default scenario registry
• In a multi-API REPL session, typing .scenario then pressing Tab suggests configured API groups
• In a multi-API REPL session, typing .scenario billing then pressing Tab suggests only billing scenarios (and not scenarios from other groups)
• In a multi-API REPL session, typing .scenario billing pets/orders/p then pressing Tab suggests nested scoped matches (for example, pets/orders/pending)
• In a multi-API REPL session, typing .scenario unknown then pressing Tab does not crash and yields no completions

Tasks

• Refactored .scenario completion logic to support multi-API staged completion (group -> scenario path/function)
• Extended createCompleter to accept grouped scenario registries for multi-runner completion routing
• Wired grouped scenario registries from REPL startup into completer initialization
• Added/updated REPL tests for:
  • group suggestions after .scenario
  • group-scoped scenario suggestions after .scenario <group>
  • nested path completion within selected group
  • unknown-group graceful fallback
  • preserved single-runner completion behavior
• Updated REPL docs for .scenario tab-completion behavior in single- vs multi-API sessions
• Added patch changeset for user-facing completion behavior update