Change LLRF:AI0-[MAG/ANG] -> LLRF:REFCOMP-CTRLIN-[MAG/ANG]
Change LLRF:AI1-[MAG/ANG] -> LLRF:REFCOMP-REF-[MAG/ANG]

Now if the a and b values are out the limit they will be set
with maximum value.

So is possible to change the values

The raw data should be converted to unsigned 16 before being converted
to unsigned 32, so the value use is correct.
This fixes avoid values greater then the maximum for the magnitudes (on
downsampled channels and PI ERROR Channel)

The higher limit is 1.999984741211, because the format for this data is
UQ1.15

Fixed the tests too.

When enable/enable-transf is changed XAxis should be updated, otherwise
it could be wrong.

Now firmware 0.5.x is not supported

Table mode was implemented just for FF and wasn't correctly.
Now is implement for both (SP and FF), and using the right registers and bits

NEARIQN is not available anymore

nearIQM and nearIQN shouldn't be change after IOC boot, so removed the PVs
and code which allow it. Change the way filters receive this parameter, now
they receive on constructor, and remove it from ControllerChannelGroup

And tests for it.

Include AI ENBL-TRANSF and remove PV aligment from SMNM of aux channels

The aligment should happen only on memory.

Lower limit is 1 and higher is AI_NELM.

So on OPI is possible to split I and Q for labels