Disable stationconstraint for fast solve when calibration is phase-only

This MR disables the stationconstraint for fast solves when the calibration is phase-only (unless the data are LBA data where there is typically not enough S/N to do so), allowing each station to get its own, individual, phase corrections. The stationconstraint is still used for the fast solve when the slow gain solves are done as well, as the phase component of the second slow solve provides the phase corrections for each station (since no stationconstraint is used for the second slow solve). Test have shown that this strategy produces better images for the phase-only cycles (see RAP-193).