Port compiler.rs operator-text helpers (compound_binop, has_empty_operand) to specs/compiler/operator_text.t27 through t27o #9109

Second step of the t27o dogfood run on bootstrap/src/compiler.rs (see #8998, see #5980). The first step is #9090. This one falls under the owner's standing rule of 2026-10-11 for ports, which covers FROZEN_HASH.

tri t27o --check converts two pure helpers, and its oracle passes on both:

• compound_binop ("|=" -> "|");
• CCodegen::has_empty_operand (did the renderer drop an operator's right side, as in (BOARD_NAME != )?).

This issue moves them into one generated plan, mounted with #[path].

Ready-made commit: branch port/compiler-operator-text (41285ef) holds this port, verified on the lab. It is stacked on #9093, so rebase it onto master once #9093 merges; FROZEN_HASH stays valid only if master did not touch compiler.rs in between (otherwise re-run t27c frozen-digest). The PR body is ready in the commit's evidence: corpus gen/gen-verilog/gen-rust/gen-c over 2270 specs byte-identical, a planted fault moves 245 rows, 504,205-input differential vs the hand Rust with 0 differences, cargo test 2990/9/2 same as master, compiler.rs +7 -60, check_all and check_budget exit 0.

Region lines to be added once the merge-side region check is live (see #9062).

User Scenarios

• Given x |= y in a spec, when the Verilog backend or assign_op_text asks for the binary operator, then compound_binop in the generated plan answers |.
• Given the rendered C expression (BOARD_NAME != ), when is_c_constant_expr checks it, then has_empty_operand in the generated plan answers true, and for (a != b) it answers false.

Requirements

• FR-001: specs/compiler/operator_text.t27 MUST define compound_binop and has_empty_operand with t27o's bodies, reusing tri::t27b::type_text's same for text equality.
• FR-002: The spec MUST carry tests that call each function by name, with expected values the hand Rust returned on the lab.
• FR-003: bootstrap/gen/rust/compiler/operator_text.rs MUST be unedited t27c gen-rust output, mounted by #[path]. Both hand bodies MUST be deleted and their call sites MUST call the generated module.
• FR-004: Generated output across the corpus MUST be byte-identical before and after, and cargo test --release MUST match clean master.

Success Criteria

• t27c test-report specs/compiler/operator_text.t27 reports 2 tests, 0 FAIL, 0 vacuous.
• grep -cE 'fn (compound_binop|has_empty_operand)\(' bootstrap/src/compiler.rs prints 0.
• gen, gen-verilog, gen-rust and gen-c over every tracked spec give the same bytes with t27c built from master and from the branch.
• The diff touches compiler.rs only inside compound_binop, has_empty_operand, gen_verilog_stmt, is_c_constant_expr and assign_op_text, apart from the one #[path] mount block at the top of the file.

Boundary

• specs/compiler/operator_text.t27
• bootstrap/gen/rust/compiler/operator_text.rs
• .trinity/seals/compiler_OperatorText.json
• bootstrap/stage0/FROZEN_HASH
• docs/now/2026-10-11-compiler-operator-text-t27o-port.md