TRI-27 emulator: six defects recorded as findings in the S05 specs

[dmitrii-f-t27]

The S05 specs of #988 (merged as gHashTag/t27#3581) state the TRI-27 machine and the .tbin container as the pinned emulator (src/tri27/emu at 976df51) runs them. Six of their findings are defects in the emulator itself. Until now they lived only inside the specs, so nobody could pick them up. Each sub-issue quotes the spec, links the code at the pin and names the conformance vector that reproduces it (gHashTag/t27 conformance/trinity/tri27_programs.json, replayed by tools/trinity_tri27.py).

Sub-issues:

• TRI-27 emulator: CALL pushes its own address, so every CALL/RET pair loops until the instruction budget #1165 CALL pushes its own address, so every CALL/RET pair loops until the instruction budget
• TRI-27 emulator: loader.load writes eight-byte words but run fetches four-byte units at pc * 4 #1166 loader.load writes eight-byte words but run fetches four-byte units at pc * 4
• TRI-27 emulator: tri_asm.zig and loader.load disagree by two bytes on where the code starts #1167 tri_asm.zig and loader.load disagree by two bytes on where the code starts
• TRI-27 emulator: the CONSTANTS section parser reads the section id as the constant count #1168 the CONSTANTS section parser reads the section id as the constant count
• TRI-27 emulator: the modulus 19683 is 3^9, but tri_cpu.zig and executor.zig label it 3^27 #1169 the modulus 19683 is 3^9, but tri_cpu.zig and executor.zig label it 3^27
• TRI-27 emulator: SACR takes its mode from an immediate the decoder never reads for SACR #1170 SACR takes its mode from an immediate the decoder never reads for SACR

Specs (gHashTag/t27 at 2e4093d): specs/isa/tri27_machine.t27 (FINDINGS, MODULUS_NOTE), specs/isa/tri27_bytecode.t27 (FINDINGS). Raised by the co-founders after the S05 review; nothing here changes the specs, which describe the emulator as it is.