modeling-discipline: dissolution findings are blocking — remove the advisory tier

[briansrls]

Summary

Removes the advisory tier from Practice 10's decidability table in docs/modeling-discipline.md. Authored at the operator's request to bring the keystone doc in line with the tightened reviewer prompts.

The table classified predicate dissolution as "advisory — candidate only" and walker dissolution as "advisory when the recursion is irregular." That advisory grade is the gray lane: a reviewer notices a dissolution finding, grades it non-blocking, the PR owner waves it off — observed live on #3250 (free_monoid_non_empty predicate dissolution noticed, graded non-blocking, dismissed). It also directly contradicts the rebuilt reviewer prompts, which now treat predicate/walker dissolution as blocking with no nit/advisory channel.

Changes (1 file, +16/-6)

• Decidability intro — every dissolution finding is blocking; no advisory tier, no nit channel. Resolved only by 🔴 dissolve-now / tracked 🟡 / substantiated 🟢 — never a free-text "intentional"/"short-circuiting" dismissal. The decidable? column now records only who flags a finding (checker vs reviewer judgment); both block.
• walker dissolution row — blocking; genuinely-irregular recursion (call graph ≠ data graph) is a clean 🟢, not an advisory.
• predicate dissolution row — blocking; a match that is genuinely distinct work is a clean 🟢. No advisory/candidate tier.
• Adds the free_monoid_non_empty coproduct-discriminant shape as the canonical predicate-dissolution example.

Notes for review

• @still-hawk-102 — this edits your keystone (Practice 10). The change is surgical (the decidability subsection + one example sentence); I did not restructure or renumber any Practice. Operator mentioned a "Practice 7" framing — left as a review question rather than guessing at a structural change.
• Operator (briansrls) authored-by-request and will review/merge.

Test plan

• Docs-only; no code/tests.
• Operator + still-hawk review.

🤖 Generated with Claude Code