Jenne noticed in alog 20099 that PR3 jumps on lock-lass when the integrators are cleared, but then comes back with a 4min time constant to more or less the same location.

Thus I implemented a split integrator for PR3 PIT:
FM1 z4min:p0  zpk([6.6315e-04],[0],6.6315e-4,"n")
FM2 :p4min    zpk([],[6.6315e-4],1507.955,"n")

Instead of clearing the filter module, the down script now simply turns off FM1 and turns it back on, producing a 4min exponential decay.

The guardian is updated to use this (set the right filters, use new clearing procedure in DOWN state).
Additionally I took out take out any clearing in LOCK_DRMI_1f. (PR3 is no used in DRMI ASC, so there should not be anything to clear anyway.)