gh-148022: Add threat model to remote debugging docs (#148024)

The remote debugging protocol has been generating spurious
vulnerability reports from automated scanners that pattern-match
on "remote access" and "memory operations" without understanding
the privilege model. This section documents the security boundaries
so reporters can self-triage before submitting.

The threat model clarifies three points: attaching requires the
same OS-level privileges as GDB (ptrace, task_for_pid, or
SeDebugPrivilege), crashes caused by reading corrupted target
process memory are not security issues, and a compromised target
process is out of scope. A subsection explains when operators
should use PYTHON_DISABLE_REMOTE_DEBUG for defence-in-depth.