Add cpu/fpu/roundmode, fpu_branch and cpu/lsu/llsc to tests_good

fpu_branch_hazard (the compare-to-branch hazard), cacheop (the write-back
data cache seen through the uncached mirror) and fpu_nan (which NaN 0/0
makes, host dependent on x86) go to tests_next. cpu/fpu/fpu is re-recorded
from a binary built with the current toolchain, which prints -nan.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

test.py

@@ -103,8 +103,11 @@ tests_good = [
   "cpu/vfpu/vavg",
   "cpu/icache/icache",
   "cpu/lsu/lsu",
+  "cpu/lsu/llsc",
   "cpu/fpu/fpu",
   "cpu/fpu/rounding",
+  "cpu/fpu/roundmode",
+  "cpu/fpu/fpu_branch",
 
   "audio/atrac/addstreamdata",
   "audio/atrac/atractest",
@@ -482,6 +485,9 @@ tests_next = [
   "cpu/vfpu/minmax_zero",  # signed zero and denormals in vmin/vmax
   "cpu/vfpu/specials",  # vcmp on denormals, NaN canonicalization and denormal flush in vbfy/vocp/vavg/vfad/vsocp
   "cpu/vfpu/overlap_vcrsp",  # vcrsp overlapping its source, which the assembler refuses; the hardware doesn't read-before-write
+  "cpu/fpu/fpu_branch_hazard",  # a bc1x right after c.xx.s sees the old condition; the compiler pads for it, not emulated
+  "cpu/fpu/fpu_nan",  # 0/0 and inf-inf give 0x7fc00000; x86 hosts make 0xffc00000, and a check per op isn't worth it
+  "cpu/lsu/cacheop",  # the data cache is write-back and the uncached mirror shows it; not emulated
   "cpu/vfpu/prefixes",
   "cpu/vfpu/vector",
   "cpu/vfpu/vregs",