docs: trusted-code assumptions

[Fitznik]

What ❔

• Document the trusted-code assumptions for all delegation circuits in docs/delegation_circuits.md, clarifying why zero control masks are allowed in the proving circuit, how the simulator enforces stricter behavior during development, and why this is safe for current M-mode firmware.
• Fix unsigned DIV/REM witness fallback in cs/src/machine/ops/mul_div.rs so that the default remainder for divisor=0 mirrors the dividend

Why ❔

Delegation circuits share the same trusted-code surface as the base machine. The new documentation explains the security model for external operators, the difference between simulator/runtime behavior, and the role of trusted firmware. This prevents confusion when reading circuit code or comparing simulator panics with circuit permissiveness.

Is this a breaking change?

• Yes
• No

Checklist

• PR title corresponds to the body of PR (we generate changelog entries from PRs).
• Tests for the changes have been added / updated.
• Documentation comments have been added / updated.
• Code has been formatted.