fix(fs): avoid recursive semantic refresh on resource delete

[fujiajie666]

Description

Deleting a resource folder enqueues a semantic refresh for the parent directory so
the parent's overview reflects the deleted child. However, _enqueue_delete_refresh()
created the SemanticMsg without overriding its default recursive=True.

That caused a delete under viking://resources to recursively re-process unrelated
sibling folders, creating thousands of unnecessary embedding tasks, API 429s, health
check timeouts, and token waste.

This PR makes delete refresh messages non-recursive, so only the parent directory is
refreshed while existing child/sibling abstracts are reused.

Related Issue

Fixes #2959

Type of Change

• Bug fix (non-breaking change that fixes an issue)

Changes Made

• fs_service.py: set recursive=False when enqueueing delete refresh SemanticMsg
• tests: assert resource delete refresh messages are non-recursive

Testing

• I have added tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes
• I have tested this on the following platforms:
  • macOS

Local verification:

uv run --extra test pytest -q tests/service/test_fs_service.py