azure-prepare skill asks "storage queue vs service bus" when user says "queue" — selection.md maps queue→servicebus

## Bug

When a user says **"create a queue listener app"**, the azure-prepare skill prompts them to choose between **Azure Storage Queue** and **Azure Service Bus**. This should not happen.

## Expected Behavior

Per `selection.md`, the intent keywords `"queue"` and `"message"` map directly to the `servicebus` resource. The skill should select Service Bus automatically without asking.

Additionally, there is **no Storage Queue recipe or resource type** in the template system at all — `functions_template_get` returns no storage queue templates. So offering Storage Queue as an option is doubly wrong: it contradicts selection.md AND there is no template to back it up.

## Actual Behavior

The skill presents a choice dialog asking the user to pick between Storage Queue and Service Bus, even though:
1. `selection.md` maps `"queue"` → `servicebus` (not ambiguous)
2. No `storagequeue` resource type exists in the MCP template catalog
3. The user never mentioned "storage" — they just said "queue"

## Steps to Reproduce

1. Invoke the azure-prepare skill
2. Say: "create a queue listener app in typescript"
3. Observe: skill asks user to choose between Storage Queue and Service Bus

## Environment

- GitHub Copilot CLI
- azure-prepare skill via MCP tools
- `functions_template_get` API

## References

- `selection.md`: maps `queue`, `message` → `servicebus`
- `composition.md`: only references servicebus recipe for queue scenarios
- MCP `functions_template_get(typescript)`: returns `servicebus-trigger-typescript-azd` but no storage queue template

Read selection.md and composition.md, then reproduce the azure-prepare flow with “create a queue listener app in typescript.” Check the intent mapping and available functions_template_get results, including the servicebus recipe and absence of a storage queue template. Done means the queue request selects Service Bus without presenting a Storage Queue choice.

Written by the indexing model from the issue text.