fix: send priming event in GET handler to prevent hang

[wukai9203]

Problem

GET requests to /mcp endpoint hang indefinitely when using streamable_http_app().

Related issue in FastMCP: jlowin/fastmcp#532

Root Cause

The _handle_get_request method's standalone_sse_writer function doesn't send a priming event before entering the message loop, unlike the POST handler which calls _maybe_send_priming_event() at line 592.

This causes the SSE writer to block waiting for messages that may never arrive, making the GET request appear to hang.

Solution

Send an event immediately after establishing the SSE connection in standalone_sse_writer():

1. With event_store and protocol >= 2025-11-25: Send proper priming event via _maybe_send_priming_event() (for SSE resumability support)
2. Without event_store or older protocol: Send a simple {"event": "open", "data": ""} event to confirm connection is established

Additional fixes:

• Add anyio.ClosedResourceError handling for clean shutdown when close_standalone_sse_stream() is called
• Store SSE writer reference so close_standalone_sse_stream() can properly close it

Testing

• Added test case test_get_request_receives_priming_event_with_event_store to verify the fix
• Manually tested with curl and verified GET requests return immediately:

  # Before fix: hangs indefinitely
  # After fix: returns "event: open\ndata: \n\n" immediately
  curl -N http://localhost:8003/mcp \
    -H "Accept: text/event-stream" \
    -H "mcp-session-id: <session_id>"

Checklist

• Code follows project style guidelines
• Tests added for the fix
• Manually tested the fix with FastMCP integration