fix(resources/prompts): list correctly; add resource read, prompt get, and Type column

[dariushoule]

Summary

mcp-cli resources and mcp-cli prompts always print "No resources available." /
"No prompts available." even when the server exposes them; there is no way to read
a resource or fetch a prompt; and the resources table's Type column always shows
unknown. This PR fixes all of it.

1. Listing always empty (root cause)

ToolManager.list_resources / list_prompts were synchronous stubs that returned the
un-awaited coroutine from the (async) StreamManager and never converted the result, so
the commands (which await them and access resource.id / prompt.name) always saw
nothing — and would AttributeError on a dict once data was available. Contrast
get_all_tools, which already awaits and converts to ToolInfo.

Fix: make both async, await the StreamManager, and convert each raw dict to a typed
object — ResourceInfo.from_raw (existed) / a new PromptInfo model.

2. Read a resource / fetch a prompt

ToolManager.read_resource existed but had no CLI/chat surface, and there was no
get_prompt path at all.

Fix:

• mcp-cli resources --read <uri> — print a resource's contents.
• mcp-cli prompts --get <name> — fetch and render a prompt's messages (adds
  ToolManager.get_prompt; the prompts command already declared a get parameter that was
  never implemented).

(Both also work as the read / get parameters in chat and interactive modes.)

3. Type column always "unknown"

The command read resource.type or resource.extra.get("mime_type"), but MCP resources use
mimeType (and are identified by uri).

Fix: ResourceInfo.from_raw normalizes uri → id and mimeType → type (originals
preserved in extra), so the Type column is populated.

Tests

• ToolManager.list_resources/list_prompts/get_prompt rewritten/added to the async + typed
  contract (the previous list tests asserted the buggy synchronous passthrough).
• ResourceInfo.from_raw normalization cases + new PromptInfo unit tests.
• resources --read and prompts --get command tests (success + not-found).
• Existing command tests already mocked list_resources/list_prompts as async returning
  ResourceInfo objects — this aligns the implementation with that contract — and still pass.
• Affected suites (tests/tools + tests/commands/definitions) all green. (Unrelated
  pre-existing Windows-console charmap failures in chat-banner tests are untouched.)

Companion fix

End-to-end output depends on chuk-tool-processor returning the data: its stdio/HTTP
transports discard the pydantic *Result models, and StreamManager lacked get_prompt.
Both are fixed in IBM/chuk-tool-processor#45. With that plus this PR, mcp-cli resources
and prompts list, read/get, and show types correctly.