core/tools: show images instead of describing them

New `view_image` reads a local picture (JPEG, PNG, GIF, WebP, sniffed from its
bytes) and hands it to the model as an image part, with an optional `detail`
level — `low` measured 3.4x cheaper on a 1200x900 picture. `read_file` now
points at it instead of returning decode noise.

`fetch_local` returns an image response as the picture itself, and MCP image
content blocks become image parts instead of `[image: ...]` placeholders, so a
screenshot server finally reaches the model.

tests/test_tools/test_catalog.py

@@ -59,6 +59,7 @@ def test_read_only_classification() -> None:
     read_only = {t.name for t in default_tool_definitions(surface="local") if t.tags & ToolTag.READ_ONLY}
     assert read_only == {
         "read_file",
+        "view_image",
         "get_truncated",
         "glob_paths",
         "regex_files",