Provide the Prim module hierarchy and desugar unary minus #89

Part of #75 (Phase 2). The last two M2 blockers, neither of which is a lowering gap.

1. The built-in Prim module hierarchy

failing/2197-shouldFail.purs and failing/2197-shouldFail2.purs both import Prim (Number) / import Prim (Boolean). We report ModuleNotFound instead of the expected ScopeConflict and UnknownName, because Prim does not exist as a source module.

purs provides Prim, Prim.Boolean, Prim.Int, Prim.Number, Prim.Row, Prim.Symbol, Prim.Ordering as compiler-provided modules. Seven passing files import Prim.Row and four import Prim.Symbol, so this is not only a failing-suite concern.

The tension to resolve: our commit 23c4fce represents primitives as constructors on the type spine rather than as a Prim module. Prim has to be provided as a module that re-exports those spine constructors, not by reintroducing a parallel primitive representation. CannotDefinePrimModules (2 corpus cases) means a user module named Prim must be rejected — so this needs a reserved-name rule too.

2. Unary minus

failing/2109-negate.purs is x = -5 annotated @shouldFailWith UnknownName. purs desugars unary minus into a call to negate, which is then an unknown name. We report nothing at all, so the case cannot agree.

Desugaring -e to negate e is a source-to-source rewrite that belongs with the other desugaring (694a3f8 lowered do to bind/discard/let). It needs negate to exist in Prelude — see Phase 3.

Also in this slice

The remaining transitive export rules, all measured M2 failures with no diagnostic today:

• RequiredHiddenType.purs → TransitiveExportError: a type hidden by the export list.
• TransitiveKindExport.purs → TransitiveExportError: a kind signature must be exported.
• value-type-based transitive exports (value referenced in an exported signature).

Acceptance

Both 2197-* cases agree, 2109-negate.purs agrees, M2 reaches 70/70, and passing files importing Prim.* resolve.