Upstream GUI/agent-friendly fixes to Mole (engine patch strategy)

[caezium]

Summary

Burrow needs a few behaviors mo doesn't expose well for a GUI/agent host.
Strategy: upstream-first — every fix is a PR to
tw93/Mole. Carry a local patch only until it
merges. Stand up a fork only if a needed patch is declined, kept minimal.

Candidates to upstream

• mo analyze --json should not open /dev/tty — fixes Burrow v0.6.5 analysis tab fails to launch mo analyze from the GUI. #35
  (could not open a new TTY: open /dev/tty: device not configured from a
  GUI process with no controlling terminal).
• Additional non-interactive / machine-readable flags as GUI-hostile
  behavior surfaces.

Process

1. Open the fix as a PR to tw93/Mole.
2. If merged → bump the pinned bundle (see the engine-bundling issue). Done.
3. If not yet merged → carry a minimal build-time patch against the pinned
   binary; keep it documented and auditable (ties to SECURITY.md).
4. Only if a needed patch is rejected outright → a caezium/Mole fork tracking
   upstream, minimal patch series, clearly attributed.

Why this way

• Contribution > competition: best answer to the "copying" framing, and it
  costs less than maintaining a fork.
• The fork is a build input of last resort, not a product.

References

• Burrow v0.6.5 analysis tab fails to launch mo analyze from the GUI. #35 — the concrete first candidate

[caezium]

I have been blocked by @.tw93 so I cannot contribute to mole