QEC: implement magic-state-consuming logical T and CCZ operations

[arniber21]

What to build

Add a scoped logical non-Clifford path for QEC-backed programs by modeling magic-state-consuming logical T/Tdag and CCZ operations in the first slice.

This should be implemented as compiler-visible operations with explicit resource accounting and QEC experiment metadata. It does not need to implement a full distillation factory in the first slice, but it must make magic-state consumption explicit rather than hiding non-Clifford behavior as generic gates.

Acceptance criteria

• Source-level syntax, intrinsic calls, or standard-library functions exist for logical T/Tdag.
• Source-level syntax, intrinsic calls, or standard-library functions exist for logical CCZ.
• Type checking enforces compatible QEC block families, code distances, operation arity, and classical/control dependencies where applicable.
• Lowering records magic-state consumption, logical non-Clifford operation kind, and affected logical IDs in the QEC workload.
• Resource reports include T count, Tdag count, CCZ count, and magic-state demand.
• QEC experiment JSON exposes the logical non-Clifford operations and their assumptions.
• Unsupported families or missing magic-state resources fail with actionable diagnostics.
• At least one T example and one CCZ example compile through the neutral-atom target and emit schedule, QEC experiment, and resource artifacts.
• Documentation states that this is a compiler model of magic-state consumption, not a validated distillation factory or threshold claim.

Blocked by

None - can start immediately.