TS => MJS: types/btree/find - #1470

Approving. Clean migration, and the BREAKING CHANGES label is correct here — worth saying explicitly, because it's the opposite call from #1468 and both are right.

There, the renamed types were declared without export in a still-TypeScript file, so nothing could observe the rename. Here all six genuinely were public:

export type FirstLeaf1<T> = readonly[Index<3>, Leaf1<T>]
export type FirstBranch3<T> = readonly[1, Branch3<T>]
export type FirstLeaf2<T> = readonly[Index<5>, Leaf2<T>]
export type FirstBranch5<T> = readonly[1|3, Branch5<T>]
export type PathItem3<T> = readonly[0|2, Branch3<T>]
export type PathItem5<T> = readonly[0|2|4, Branch5<T>]

so demoting them to _-prefixed removes names a consumer could have imported. Breaking, and labelled as such.

The demotion itself is justified. I checked whether anything outside btree/find depends on them, since a _-prefixed name that another module imports would contradict the convention. Nothing does — the only hits in fjs/types/btree/set/module.f.ts are lines 52-56, a commented-out block illustrating the type shapes, not real imports. btree/set imports only find, First, PathItem and Result, all of which stay public.

That's the right cut: First remains the public union, and the four constituent shapes plus the two PathItem variants become private detail. A consumer that needs to name a branch of First can still narrow on it structurally.

Retained exports verified — find, isFound, value, First, Path, PathItem, Result all import cleanly from the new .mjs, with each generic instantiated at <1> so the type parameters are checked too.

Proper add + delete pair. npx tsc clean, npm test → pass: 2355, fail: 0, total: 2355, CI 19/19.

Very minor, not worth a commit on its own: the commented block at btree/set/module.f.ts:52-56 still transcribes First as FirstLeaf1<T> | …, which is now _FirstLeaf1<T> | …. It reads as an explanation of the shape rather than a pointer to importable names, so it's still accurate as prose.