Gimlet Sequencer doesn't correctly snapshot fault state/pg

[Aaron-Hartwig]

While debugging a unit that fails to sequence, the logic for when to snap FLT_A0_SMSTATUS, FLT_GROUPB_PG, and FLT_GROUPC_PG appears to be incorrect as they stay at zero. We can see that the watermark (MAX) registers are latched, but not the fault ones

$ humility ringbuf gimlet_seq
...
  42  950        1        1 A0Failed(A0Timeout)
  43  940        1        1 A0FailureDetails(IFR, 0x0)
  44  940        1        1 A0FailureDetails(DBG_MAX_A0SMSTATUS, 0xa)
  45  940        1        1 A0FailureDetails(MAX_GROUPB_PG, 0xff)
  46  940        1        1 A0FailureDetails(MAX_GROUPC_PG, 0x3)
  47  940        1        1 A0FailureDetails(FLT_A0_SMSTATUS, 0x0)
  48  940        1        1 A0FailureDetails(FLT_GROUPB_PG, 0x0)
  49  940        1        1 A0FailureDetails(FLT_GROUPC_PG, 0x0)
...