I succeeded in switching PRCL over to REFL_A_RF45_I.

This improved the sensing noise between 30Hz and 100Hz by about a factor of 2.

We did this in the middle of an earthquake, and eventually lost lock, so I didn't put it in Guardian yet. We also had more than usual scattering due to the earthquake, so it is too early to tell whether it made a DARM noise difference.

Also:

- We should investigate increassing (or better: not reducing) the RF45 modulation index.

- By mixing POP_A_RF9_I and REFL_A_RF45_I to PRCL by about 0.26:0.74 we should get another ~25% shot noise reduction.

- The first lock looked promising, but if for some reason we still have a DC drift problem on REFL, we could blend it to POP_A_RF9_I at DC

Matrix elements:

replaced the old

POP_A_RF9_I to PRCL = 0.035

with the new

REFL_A_RF45_I to PRCL = -1147.5 (yes, a minus sign)