Gaincal access to visibilities speed up

Accessing the visibilities using casacore indexes causes a very expensive operation in the ok() method. Such method, among other things, checks the integrity of the array at each access.

Since the array is consecutive the pointer can be used to create a span speeding up the performances of a factor ~2.

This is the DP3 before the optimization

Total DP3 time     119.93 real     2611.46 user      118.48 system
    1.1% ( 1351 ms) MSReader
   95.5% (  114  s) GainCal gaincal.
           44.8% (   51  s) of it spent in predict
           51.9% (   59  s) of it spent in reordering visibility data
            2.3% ( 2591 ms) of it spent in estimating gains and computing residuals
            0.0% (   30 ms) of it spent in writing gain solutions to disk
        Converged: 91, stalled: 9, non converged: 0, failed: 0
        Iters converged: 6, stalled: 9, non converged: 0, failed: 0
    2.9% ( 3467 ms) MSWriter msout.
      3.4% (  119 ms) Creating task

and this is after the optimization

Total DP3 time      65.06 real      964.02 user      133.96 system
    2.1% ( 1344 ms) MSReader
   92.2% (   59  s) GainCal gaincal.
           90.4% (   54  s) of it spent in predict
            2.6% ( 1553 ms) of it spent in reordering visibility data
            4.5% ( 2684 ms) of it spent in estimating gains and computing residuals
            0.0% (    7 ms) of it spent in writing gain solutions to disk
        Converged: 91, stalled: 9, non converged: 0, failed: 0
        Iters converged: 6, stalled: 9, non converged: 0, failed: 0
    4.7% ( 3083 ms) MSWriter msout.
      0.3% (    8 ms) Creating task
    183.9% ( 5672 ms) Writing (threaded)