security: fix four vulnerabilities (GHSA-779p-m5rp-r4h4) (#14734)

* security: fix five vulnerabilities (GHSA-779p-m5rp-r4h4)

- CVE-2026-56670: force download of SVG/XML responses on /view to prevent stored XSS
- CVE-2026-56671: contain /experiment/models/preview reads within the model folder
- CVE-2026-56672: stop inline rendering of uploaded /userdata/{file} content
- CVE-2026-56673: prevent path traversal in get_annotated_filepath (LoadImage /prompt input)
- CVE-2026-56674: reject opaque/null Origin to close the CSRF middleware bypass

Adds regression tests under tests-unit/security_test/ covering all five.

* security: address review feedback on GHSA-779p fixes

- Fix Windows CI failure in test_get_annotated_filepath: compare against
  os.path.abspath(...) to match the intentional abspath normalization added
  by the traversal hardening (abspath prepends the drive letter on Windows).
- origin_check: narrow the bare `except:` in is_loopback() to ValueError so
  genuine interrupts aren't swallowed (review nit).
- origin_check: guard .port access in is_cross_origin_forbidden() so a
  malformed/out-of-range port (e.g. Origin: http://127.0.0.1:99999) fails
  closed with a 403 instead of surfacing an uncaught 500 in the middleware.
- server /view: escape backslash/quote in the Content-Disposition filename
  (RFC 6266 quoted-string) so a filename containing a double quote can't
  malform the response header.

* security: address CodeRabbit review feedback on GHSA-779p tests

- test #3: guard the symlink-escape test with a try/except skip so it no
  longer errors on Windows CI where os.symlink needs elevated privileges /
  Developer Mode (mirrors the guard in the sibling test #2).
- test #5: refresh the stale module docstring to describe the actual /view
  gating (view_image closure calling folder_paths.is_dangerous_content_type,
  the normalising check) instead of the bypassable raw set-membership test.

* revert(security): drop CVE-2026-56674 Origin: null CSRF change

Per maintainer review, the reported CSRF is already mitigated by the pre-existing
Sec-Fetch-Site: cross-site check for current browsers, and the null-origin
rejection risked breaking legitimate sandboxed-iframe embeds. Restores
origin_only_middleware and is_loopback in server.py to their prior state
(the Sec-Fetch-Site check is retained) and removes utils/origin_check.py and its
regression test. The other four GHSA-779p fixes are unaffected.

folder_paths.py

@@ -264,6 +264,59 @@ def annotated_filepath(name: str) -> tuple[str, str | None]:
     return name, base_dir
 
 
+# Content types a browser may execute or render inline. File endpoints that
+# serve user-controlled content must force these to download (and ideally set
+# Content-Disposition: attachment) to avoid stored XSS. Centralised here so the
+# /view and /userdata handlers can't drift apart. mimetypes.guess_type may
+# return either the text/* or application/* spelling depending on platform, so
+# both are listed.
+DANGEROUS_CONTENT_TYPES = {
+    'text/html', 'text/html-sandboxed', 'application/xhtml+xml',
+    'text/javascript', 'application/javascript', 'application/x-javascript',
+    'application/ecmascript', 'text/css',
+    'image/svg+xml', 'application/xml', 'text/xml',
+    # message/rfc822 (.mht/.mhtml) can carry script in some browsers.
+    'message/rfc822',
+}
+
+
+def is_dangerous_content_type(content_type: str | None) -> bool:
+    """Return True if a browser may execute or render `content_type` inline.
+
+    Normalises before matching so the check can't be slipped past with a
+    charset/boundary parameter (``text/html; charset=utf-8``) or casing
+    (``TEXT/HTML``). Any XML dialect (``*+xml`` or ``*/xml``) is treated as
+    dangerous because XML can carry inline script via stylesheet/entity tricks,
+    which also covers the ``application/{xslt,rss,atom,rdf}+xml`` family without
+    enumerating each one. Endpoints serving user-controlled content should route
+    a dangerous type to ``application/octet-stream`` + ``Content-Disposition:
+    attachment`` + ``X-Content-Type-Options: nosniff``.
+    """
+    if not content_type:
+        return False
+    normalized = content_type.split(';', 1)[0].strip().lower()
+    if normalized in DANGEROUS_CONTENT_TYPES:
+        return True
+    return normalized.endswith('+xml') or normalized.endswith('/xml')
+
+
+def is_within_directory(directory: str, target: str) -> bool:
+    """Return True if `target` resolves to a path inside `directory`.
+
+    Uses realpath on both operands so that a symlink placed inside `directory`
+    that points elsewhere cannot escape the containment check at open time.
+    """
+    try:
+        directory = os.path.realpath(directory)
+        target = os.path.realpath(target)
+        return os.path.commonpath((directory, target)) == directory
+    except ValueError:
+        # ValueError is raised by realpath() on a path with an embedded null
+        # byte, and by commonpath() on Windows when the paths are on different
+        # drives. In either case the target is not safely within the directory.
+        return False
+
+
 def get_annotated_filepath(name: str, default_dir: str | None=None) -> str:
     name, base_dir = annotated_filepath(name)
 
@@ -273,7 +326,12 @@ def get_annotated_filepath(name: str, default_dir: str | None=None) -> str:
         else:
             base_dir = get_input_directory()  # fallback path
 
-    return os.path.join(base_dir, name)
+    filepath = os.path.abspath(os.path.join(base_dir, name))
+    # Prevent path traversal: the resolved path must stay within base_dir.
+    # repr() the name in the message so a crafted value can't inject log lines.
+    if not is_within_directory(base_dir, filepath):
+        raise ValueError("Invalid file path: {!r}".format(name))
+    return filepath
 
 
 def exists_annotated_filepath(name) -> bool:
@@ -282,7 +340,10 @@ def exists_annotated_filepath(name) -> bool:
     if base_dir is None:
         base_dir = get_input_directory()  # fallback path
 
-    filepath = os.path.join(base_dir, name)
+    filepath = os.path.abspath(os.path.join(base_dir, name))
+    # Treat traversal attempts as non-existent rather than probing the filesystem.
+    if not is_within_directory(base_dir, filepath):
+        return False
     return os.path.exists(filepath)