P102: the store buffer is not correct yet

P102 tried to add a core-local store buffer after P101 split the TLB. Verilator builds, Linux boots, and /init starts. BusyBox then faults before the shell prompt.

The useful data point is the direct +stop_on_kernel_panic run:

accepted stores : 79
drained stores  : 79
forwards        : 0
valid cycles    : 79
drain stalls    : 0
panic milestone : 118,576,336 cycles

The panic is deterministic:

init[1]: unhandled signal 11 code 0x1 at 0xffffff88 in busybox[a9d42,10000+117000]
epc : 000b9d42
badaddr: ffffff88 cause: 0000000d

0x000b9d42 is lw a4,-120(s0), and the trap dump has s0 = 0. So this is not a vague performance regression. It is a concrete correctness bug around early BusyBox register/stack/global state.

Honest status: PARTIAL. The RTL builds, but the BusyBox shell smoke is FAIL. Next work should trace the first buffered user stores instead of layering more nonblocking memory machinery on top.