docs: gate prototypes on persistent uncertainty

[iteathen]

Summary

• clarify that low confidence after the first assess -> research -> reassess cycle does not immediately trigger a prototype
• require another deeper, broader, or more targeted research pass followed by reassessment
• introduce a three-state confidence gate: sufficient confidence -> final plan; low confidence with productive research -> research again; persistent low confidence after additional research / research saturation -> rapid prototype
• keep prototypes as reconnaissance for uncertainty that research is no longer resolving efficiently
• update the governing flow accordingly

Scope

Documentation/governance only. No runtime, schema, ABI, support, or compatibility claims change.

Validation

Protected main requires the Community quality check. Merge only the exact reviewed head after it passes.