nx_rv64_toolchain_gate.nx

nx_rv64_toolchain_gate.nx -- END-TO-END SOVEREIGN TOOLCHAIN CLOSURE: prove the assembler and the emulator are now ISA-CONSISTENT (both RV64IM) by running a REAL multiply/divide algorithm through the WHOLE pipeline -- text asm --(nx_rv64_asm)--> machine code --(fk_predecode)--> decoded --(tier_run: interpret cold + JIT hot)--> result. The program (sum of squares 1..10 = 385, then /7 and %7) uses MUL in a hot loop (JIT-compiled) + DIV/REM (tier to interp) -- exercising exactly the M extension just added to BOTH the assembler and the emulator. Closes the debt that the assembler emitted only RV64I while the emulator executed RV64IM (the M-ext gates set FK_ constants directly, bypassing the assembler -- this gate does NOT). T1 assembler M-ext ENCODING KAT: 'mul x8,x6,x6' assembles to the exact RISC-V word 0x02630433. T2 end-to-end on the golden interpreter: assembled program computes 385 / 55 / 0. T3 end-to-end on the TIERED JIT engine: same result, and the MUL loop was actually JIT-compiled + run native. T4 the two agree bit-for-bit (the assembler->tiered-JIT path == the golden path). expect_exit: 0