P118 - LSU shape counters

P118 adds counters that describe the current monolithic data path as a future load/store unit. No behavior changed: loads, stores, AMOs, FP memory ops, DTLB checks, D-cache hits, store-buffer accepts, and lower-memory handshakes still flow through execute plus S_MEM.

The shell profile is PASS:

post-load cycles    : 218,857,370
shell window cycles : 64,957,904
retired instructions: 86,383,280
CPI                 : 2.5336
S_FETCH cycles      : 7,628,220
S_MEM cycles        : 27,758,933

The new LSU-shaped counters are the useful part:

addrgen total      : 27,899,292
addrgen loads      : 15,536,292
addrgen stores     : 12,161,761
addrgen AMOs       : 201,239
DTLB hits          : 27,223,125
DTLB misses        : 671,678
S_MEM LSU cycles   : 27,758,933
D-cache hit cycles : 4,603,337
storebuf accepts   : 1,156,649
storebuf waits     : 0
aux load cycles    : 0

Honest status: RTL simulation PASS. LibreLane hardening is NOT RUN.