Task.read has inconsistent return type between normal AI and power channels

[bkeryan]

For normal AI channels, leaving samples_to_read unset reduces the dimensions of the resulting list:

• Single channel task:
  • task.read() -> scalar
  • task.read(1) -> 1D list by sample
  • task.read(N) -> 1D list by sample
• Multi-channel task:
  • task.read() -> 1D list by channel
  • task.read(1) -> 2D list by channel, sample
  • task.read(N) -> 2D list by channel, sample

However, for power channels, explicitly setting samples_to_read=1 has the same effect:

• Single channel task:
  • task.read() -> scalar
  • task.read(1) -> scalar
  • task.read(N) -> 1D list by sample
• Multi-channel task:
  • task.read() -> 1D list by channel
  • task.read(1) -> 1D list by channel
  • task.read(N) -> 2D list by channel, sample

Passing samps_per_chan=1 to task.read() returns a different data type, which may cause errors for programs that don't expect that to happen.